Gender Discrimination and Economic Outcomes in Chile

GENDER DISCRIMINATION AND ECONOMIC OUTCOMES IN CHILE Final Report List of Authors: David Bravo Director, Centro de Microdatos, Departamento de Economía Universidad de Chile dbravo@econ.uchile.cl Claudia Sanhueza Investigadora Asociada, Centro de Microdatos, Departamento de Economía Universidad de Chile csanhueza@econ.uchile.cl Sergio Urzúa Post Doctoral Student, University of Chicago surzua@uchicago.edu November 29, 2006 Table of Contents Presentation ………………………………………………………………………… 2 Chapter 1: “An Experimental Study about Labor Market Discrimination: Gender, Social Class and Neighborhood”………………………………………………………………… 3 Chapter 2: “Ability, Schooling Choices and Gender Labor Market Discrimination: Evidence for Chile”…………………………………………………………………………… 35 Chapter 3: “Is there labor market discrimination among professionals in Chile? Lawyers, Doctors and Business-people” ……………………………………………………… 67 1 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Presentation The study “Gender Discrimination and Economic Outcomes in Chile” is an ambitious attempt at providing systematic evidence for an area of great interest. This project is comprised of three studies that the Centro de Microdatos of the Department of Economics of the Universidad de Chile is developing in parallel. This final report presents the final papers in its three components. Firstly, the chapter “An Experimental Study about Labor Market Discrimination: Gender, Social Class and Neighborhood” is presented. This paper examines whether there are differences in the response rate for curriculum vitae sent in response to job vacancies published in the main newspaper of Chile, both for gender and socioeconomic characteristics. In the version of the article presented, results are obtained for 3,500 CVs sent. Secondly, the study “Is there labor market discrimination among professionals in Chile? Lawyers, Doctors and Business-people” is presented. This paper is done based on a survey colleted specially for these effects. It is the first database of professionals in Chile containing data on non-cognitive abilities, real labor market history, social and family background and others. The aim is to investigate whether taking into account many variables that in generally are unobservable we still have differences in wages to only to gender. Finally, the third chapter of this project includes is called “Ability, Schooling Choices and Gender Labor Market Discrimination: Evidence for Chile”. This paper studies the importance of schooling decisions and abilities in explaining gender gaps in wages. In the analysis we use a rich data set for Chile (the Social Protection Survey). This data set contains information not only on labor market outcomes but also on schooling attainment and schooling performance. We introduce an empirical model in which agents make schooling and labor market choices based on this unobserved ability (in addition to observables). We use schooling performance to identify the distribution of unobserved abilities in the population. Acknowledgements The authors want to thank Verónica Flores and Bárbara Flores for their excellent research assistance and the Survey Unit of the Centro de Microdatos. 2 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Chapter 1 “An Experimental Study about Labor Market Discrimination: Gender, Social Class and Neighborhood” Abstract The objective of this chapter is to study the Chilean labor market and determine the presence or absence of gender discrimination. In order to break past the limitations of earlier works, an experimental design is used, the first of its kind in Chile. This study also allows socioeconomic discrimination associated to names and places of residence in the Chilean labor market to be tackled. The study consists of sending fictitious Curriculum Vitae for real job vacancies published weekly in the “El Mercurio” newspaper of Santiago. A range of strictly equivalent CVs in terms of qualifications and employment experience of applicants are sent out, only varying in gender, name and surname, and place of residence. The study allows differences in call response rates to be measured for the various demographic groups. In the version of the article presented, results are obtained for 6,300 CVs sent. Our results show no significant differences in callback rates across groups, in contrast with what is found in other international studies. 3 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 1 Introduction Gender and social discrimination in the labor market are some of the key issues in the discussion on public policies in Latin America. However, empirical evidence and academic research on the matter have been rather scarce until now. This is also the case in Chile. No matter how much has been done to study labor market discrimination, either racial, ethnic or gender, the issue of detection is still unsettled. In the usual regression analyses there are several problems of unobservable variables that clearly bias the results (Altonji and Blank, 1999; Neal and Johnson, 1996) and, on the other hand, experimental studies have been under discussion for not correctly measuring discrimination (Heckman and Siegelman, 1993; Heckman, 1998). In Chile, despite the fact that the average years of schooling of Chilean female workers are not statistically different from those of male workers, average wages of male workers are 25% higher 1 . In fact, previous studies 2 suggest that gender discrimination is a factor in determining wages in the Chilean labor market. Estimates of the Blinder-Oaxaca decomposition give “residual discrimination” a significant participation in the total wage gap 3 . The evidence also shows stable and systematic differences in the returns to education and to experience by gender along the conditional wage distribution. Additionally, it has been shown that “residual discrimination” is higher for women with more education and experience. Furthermore, Chilean female labor force participation is particularly low, 38.1% compared to 44.7% in Latin America 4 . This is even lower for married women and in fact, higher participation is found in separated or divorced women (Bravo, 2005). This latter fact may be interpreted as evidence of female preferences for non-market activities 5 . However, this “residual discrimination” is only a measure of how much of the wage gap is due to unobservable factors. Therefore, these measures of discrimination are biased due to the lack of relevant controls. A recent study on discrimination by social class in Chile (Núñez 1 Own calculations using CASEN 2003. Once you correct for human capital differences and occupational choice this gap falls to 19% approximately. 2 Previous studies for Chile are Bravo (2005); Montenegro (1998); Montenegro and Paredes (1999) and Paredes and Riveros (1994). 3 Bravo (2005) shows that taking all employed workers and after controlling for years of schooling and occupation, the wage gap was 13.5% in 2000. Using the Blinder-Oaxaca decomposition he concludes that most of this difference was due to “residual discrimination”. 4 Source: International Labor Organization (ILO). 5 Contreras and Plaza (2004) also found that there are cultural factors, such as sexism, that significantly influence female labor force participation in Chile. 4 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E and Gutiérrez, 2004) uses a dataset which reduces the role of unobserved heterogeneity across individuals, but it has several limitations 6 . Furthermore, there are no attempts at studying discrimination using neither audit studies nor natural experiments. The objective of this chapter is to study the Chilean labor market and determine the presence or absence of gender discrimination. In order to break past the limitations of earlier works, an experimental design is used, the first of its kind in Chile. This study also allows socioeconomic discrimination associated to names and places of residence in the Chilean labor market to be tackled. The study consists of sending fictitious Curriculum Vitae for real job vacancies published weekly in the “El Mercurio” newspaper of Santiago. A range of strictly equivalent CVs in terms of qualifications and employment experience of applicants are sent out, only varying in gender, name and surname, and place of residence. The study allows differences in call response rates to be measured for the various demographic groups. In the version of the article presented, results are obtained for 6,300 CVs sent. The following section contains a review of the relevant literature for this study. Meanwhile, Section III contains all the methodological information associated to the implementation of the experiment, which began to be applied in the last week of March 2006. Lastly, Section IV contains the preliminary results recorded to date. 2 Literature Review Labor market discrimination is said to arise when two identically productive workers are treated differently on the grounds of the worker’s race or gender, when race or gender do not in themselves have an effect on productivity (Altonji and Blank, 1999; Heckman, 1998). However, there are never identical individuals. There are several unobservable factors that determine individual performance in the labor market (see literature review in chapter 2). The empirical literature attempts to face these problems by two alternative methodologies: regression analysis and field experiments 7 . The regression analysis is focused on analyzing the Blinder-Oaxaca decomposition (Oaxaca, 1973; Blinder, 1973) to determine how much of the wage differential between groups of workers, by race or gender, is unexplained. This unexplained part is called discrimination. 6 See Section II for a discussion. See Altonji and Blank (1999) and Blank, Dabady and Citro (2004) for complete surveys on the econometric problems involving detecting discrimination in the labor market using regression analysis and field experiments. 7 5 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Developments in Chile have been centered on regression analysis applied to the gender gap. See Paredes and Riveros (1993), Montenegro (1999) and Montenegro and Paredes (1999) as an example. The conclusions from these studies are very limited. They lack several control variables, related to cognitive and non-cognitive abilities and school and family environments. In addition, preferences over non-market activities and experience of Chilean female workers could prove to be a very important unobservable factor. More recently, Núñez and Gutiérrez (2004) study social class discrimination in Chile under the traditional Blinder-Oaxaca decomposition. They use a dataset that allows them to reduce the role of unobservable factors by limiting the population under study and having better measures of productivity (see more details on these references in chapter 2). The above represent the traditional studies in this area. The present article is much more closely related to a different line of research on labor market discrimination: experimental studies 8 . These originated in Europe in the 1960s and 1970s, the ILO in the 1990s and recently experimental techniques have been published in leading economic journals (Bertrand and Mullainathan, 2004). Experimental approaches can be divided into two types: audit studies and natural experiments. The latter ones take advantage of unexpected changes in policies or events (Levitt, 2004; Antonovics, Arcidiacono and Walsh, 2004, 2005; Goldin and Rouse, 2000, Newmark, Bank and Van Nort, 1996). In Chile, as far as we know, there are no studies using these kinds of variations. There have been two procedures used to carry out audit studies. First, the personal approach strategy, which sends individuals to job interviews or does job applications over the telephone. Second, there is the strategy of sending written applications for real job vacancies. The first procedure is the most subject to criticisms. It has been argued that it is impossible to ensure that false applicants are identical. Also, testers were sometimes warned that they were involved in a discrimination study and their behavior could bias the results. 9 The first experiments that used written applications were unsolicited jobs-applications and posted to “potential employers”; these experiments tested preferential treatment in employer responses and not the hiring decision. Later came the ones that sent curriculum vitae to real solicitudes. Despite the fact that this latter technique overcomes the criticisms of the personal approaches and tests the hiring decision 10 it does not overcome a common problem of the audit studies raised by Heckman and Siegelman (1993) and Heckman (1998), which is that audits are crucially dependent on the distribution of unobserved characteristics for each 8 Riach and Rich (2002, 2004) and Anderson, Fryer and Holt (2005) have a complete survey of these studies. See Heckman and Siegelman (1993). 10 It really tests the calling back decision. We do not know what can happen next. 9 6 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E racial group and the audit standardization level. Thus, there may still be unobservable factors, which can be productivity-determining and not discrimination. Riach and Rich (2002) accepted this criticism but pointed out that it is not easy to imagine how firm internal attributes 11 could enhance productivity. They conclude that while Heckman and Siegelman (1993) do not explain what could be behind those gaps the argument has “not been proven”. The present study mainly follows the line of work developed by Bertrand and Mullainathan (2004). In their study, the authors measured the racial discrimination in the labor market, by means of the posting of fictitious curriculum vitae for job vacancies published in Boston and Chicago newspapers. Half of the CVs were randomly given Afro-American names and the other half received European names. Additionally, the effect of applicant qualification on the racial gap was measured; for this, the CVs were differentiated between High Qualifications and Low Qualifications. The authors found that the curriculum vitae associated to White names received 50% more calls for interview than those with Afro-American names. They also found that whites were more affected by qualification level than blacks. Additionally, the authors found some evidence that employers were inferring social class based on the applicants’ names. 3 Experiment Design As already indicated, the experiment consists of the sending out of CVs of fictitious individuals for real job vacancies that appear weekly in the newspaper with the highest circulation in Chile. Each week, the work team selects a total of 60 job vacancies from the “El Mercurio” newspaper of Santiago. A total of 8 CVs, 4 corresponding to men and 4 to women are sent out for each vacancy. The details of the experiment design are presented here below. 3.1 Definition of Demographic Cells We defined eight relevant demographic cells which determined the categories being studied in the experiment. Thus, eight CVs -that are equivalent as regards employment productivity but differ in the variables in question- were sent out for each vacancy. 11 Such as internal promotion or other. 7 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E The cells were defined to serve the objectives of the study. Firstly, the study of discrimination by gender requires men and women to be separated. Additionally, socioeconomic discrimination is examined by means of two variables: surnames and municipalities of residence. In order to reduce the number of observations required in each case, these last variables were separated into two groups, the most extreme: socioeconomically rich and poor municipalities; and surnames associated to the Upper Classes and Lower Classes. Since we have three dichotomous variables, the final number of demographic cells is 8 (2*2*2), as the following Table shows: Hombre Apellido Apellido Clase Alta Clase Baja Mujer Apellido Apellido Clase Alta Clase Baja Comuna Ingresos Altos Comuna Ingresos Bajos From the outset of the field work, in the last week of March 2006, each week approximately 60 job vacancies are chosen. Eight CVs are sent for each job vacancy, in other words, one for each demographic cell. So, each week 480 CVs are sent, 240 to men and 240 to women. A group of names, surnames and municipalities are established to satisfy the requirements of each cell, with the names and municipalities chosen randomly for each vacancy. 8 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E JOB VACANCY 8 Curriculum Vitae 4 Female CVs 4 Male CVs High Class Surname High Income Municipality High Class Surname High Income Municipality High Class Surname Low Income Municipality High Class Surname Low Income Municipality Low Class Surname High Income Municipality Low Class Surname High Income Municipality Low Class Surname Low Income Municipality Low Class Surname Low Income Municipality 3.2 Description of the source of job vacancies The main source of job vacancies in Santiago is the “El Mercurio” newspaper, which publishes every Sunday around 150 job vacancies, with a repeat rate of around 30%. The ads are also available in the newspaper webpage, which facilitates access (see: http://empleos.elmercurio.com/buscador/destacados/listado_destacado.asp) To prepare the field work, a prior study was carried out on this source. To this effect, in the month of January and the first three weeks of March 2006, all vacancies published were analyzed in order to design the future mailing strategy. In effect, it was possible to conclude from this study that it is necessary to prepare a CV bank based on three categories: professionals, technicians (skilled workers) and unskilled workers. Other markedly male or female categories were rejected and the approximate number of vacancies for each week was calculated. 3.3 Creation of CV Banks. As indicated above, job vacancies are grouped into three categories: Professionals, Technicians and Unskilled workers. A person was assigned responsibility for each category, and is in charge of selecting the weekly vacancies, as well as the production, sending and supervision of the CVs sent. 9 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Each person in charge (three in total) has a CV bank obtained from the www.laborum.com and www.infoempleo.cl websites. These are used as the base information for producing fictitious CVs and comply with the profile of the most competitive applicant for the vacancy selected. Each set of 8 CVs is constructed so that its qualification levels and employment experience are equivalent, in order to ensure that the applicants are equally eligible for the job in question. The central element in the training of the people in charge of this was to ensure that the 8 CVs sent for each vacancy had to be equivalent in terms of qualifications and human capital. To ensure this, the coordinators of the study were supported by a research assistant that supervised the work over the whole period and, especially, during the first weeks, until ensuring the desired results. 3.4 Classification of municipalities In order to facilitate the field work, the study is concentrated in the Metropolitan Urban Region, which is divided into 34 municipalities. To classify the municipalities in the two extreme segments, the socioeconomic classification of households based on the 2002 Census designed by Adimark was used. This institution classified from the CASEN 2003 Survey. Using this, it classifies the proportion of the population by socioeconomic level in each municipality. The groups are ordered from higher to lower level, ABC1, C2, C3, D and E. For high income municipalities, 5 of the 6 with the highest proportion of the population in segment ABC1 were chosen (the sixth was excluded because it is a municipality that also had a higher proportion than the rest in segment D and E). On the other hand, for the low income group municipalities, the 15 municipalities associated to a lower proportion of the population in segment ABC1 and a greater proportion in the segments D and E were chosen. In order to more clearly examine the impact of the municipality of origin, all other municipalities of intermediate socioeconomic groups were left out. The final list of the municipalities included in each group is presented in the Table below: 10 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Selected Municipalities HIGH INCOME MUNICIPALITIES LOW INCOME MUNICIPALITIES Pedro Aguirre Cerda Vitacura Pudahuel Conchalí Quilicura Providencia San Joaquín Lo Prado San Ramón La Reina Lo Espejo Renca Recoleta Las Condes San Bernardo La Granja Cerro Navia Ñuñoa El Bosque La Pintana 3.5 Classification and selection of names and surnames The names and surnames were classified and selected according to the procedure used by Núñez and Gutiérrez (2004). Specifically, a sample of names and surnames was taken from the alumni register of the Faculty of Economics and Business of the Universidad de Chile. Subsequently, a group of students was chosen, who classified (based on their personal perception) these names and surnames into: High Social Class, Middle Social Class and Lower Social Class. For the purposes of the field work, only the names and surnames classified as Upper Class and Lower Class were considered. An example of the surnames used in each category is presented in the Table below. SELECTED SURNAMES UPPER SOCIAL LOWER SOCIAL CLASS SURNAMES CLASS SURNAMES Rodrigo Recabarren Merino Valeska Angulo Ortiz Susan Abumohor Cassis Pablo Ayulef Muñoz Javiera Edwards Celis Rosmary Becerra Fuentes Pedro Ariztia Larrain Clinton Benaldo Gonzalez 11 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 3.6 Description of the field work The three people responsible indicated above handle the weekly selection of job vacancies that appear in “El Mercurio” on Sundays. They then construct the targeted CVs for each vacancy, with the most competitive CVs, ensuring their equivalence, in order to ensure that the only differentiating elements are the sex of the applicant, the social level, name and surname, and the municipality of residence. Apart from the people in charge, the team is made up of three other assistants. This entire procedure is supervised directly by a Sociologist and an Economist who randomly reviews the CVs sent. The job vacancies selected and the lot sent for each vacancy are entered weekly into a specially designed web page that allows all the vacancies to be reviewed, together with their respective sets of CVs. The entry of that information into the web page is handled by an I.T. expert. A central aspect of this work is receiving the calls for the CVs sent. To receive these callsresponses, there is a fully dedicated man and woman team, ready to take the calls 24 hours a day from Monday to Sunday. There are 8 mobile telephones, each with a different number, assigned to each of the CVs of the set; this ensures that the recruiters do not encounter repeated telephone numbers. The people in charge of receiving the calls record the day, name of the applicant, the vacancy and the phone number of the firm that selected the CV. Each report is entered into the web page of the project, which allows for the regular supervision of the calls received. In parallel, job vacancy responses are also received by e-mail. Some job vacancies request emails. For this, a generic e-mail has been created for each CV. To date, 54 e-mails addresses have been set up and they are checked every three days. As with the phone calls, the e-mails are reported and entered into the web page of the project. 12 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 3.7 Identity of the fictitious applicants Once the names and surnames are classified by categories, Upper Class and Lower Class, they are then mixed so as to not use real names. Additionally, each fictitious applicant has an artificial RUT 12 and an exclusive e-mail address. To ensure the equivalence of each set of CVs, the age of the applicants was set at between 30 and 35 years of age, married with between one and two children at most. 3.8 Ensuring the equivalence of the fictitious applicants between cells In order to ensure the equivalence of the 8 fictitious applications sent for each vacancy, the other variables included in the CVs were controlled for similarity. For this, the following decisions were made: • • • • 12 13 As regards the educational background of the applicants, those with university education are considered Universidad de Chile graduates and where necessary, they have postgraduate studies from the same University. The school of the applicant and the home address are determined by the municipality that they belong to. A bank of school names of each municipality is used for this. However, to ensure homogeneous schooling the 8 CVs sent must belong to the same category of socioeconomic background of the school 13 . Additionally, each CV of the set of 8 has a unique telephone number different to the other seven; however, these may repeat themselves among different groups of CVs. The employment experience of the applicants is equivalent within each category (professional, technician, unskilled worker) but different among each other. Thus, professionals with greater time spent in the educational system have less years of employment experience, meanwhile, unskilled workers have a longer track record in the labor market. To maintain this equivalence, we have also set the number of jobs that each applicant has had in the various categories and their employment history continuity (absence of employment gaps). National Identification Number. This is a discrete variable which describe the level of income of the majority of the school population. 13 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Category Professionals Technicians Unskilled workers • • • Employment Experience 7 to 12 years 8 to 13 years 12 to 17 years Number of jobs 2 to 3 4 to 5 5 to 7 Postgraduate studies of applicants are equivalent within the set of 8 CVs that are sent for each vacancy. Within the set of 8 CVs, postgraduate studies must be from the same university (Universidad de Chile) and in very similar areas or even identical areas. Training courses must also be from equivalent institutions (Technical Insitutes) and in similar or identical areas. As a general rule, high quality CVs were sent for each vacancy. In other words, the variables of employment history, education and training were drawn up to be attractive to firms. The pay expectations required, which generally have to be included in job applications, were based on actual remuneration information of professionals and technicians (from the web page www.futurolaboral.cl). The starting point were pay levels required by a good candidate (of percentile 75 of that distribution) but, subsequently, the remuneration was reduced to average levels. Each set of 8 CVs sent for a vacancy had the same reference pay level and varied only slightly (in some cases it was given as a range and in others as a specific reference). 14 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 4 Findings The CV mailing process started in the last week of March. As we can note from Table 1, on average, 68 vacancies have been applied for weekly, with a total of 11,016 CVs sent during the 20 weeks of the project, and a response rate of around 14.65%. This rate is somewhat higher than that obtained by Bertrand and Mullainathan (2004). Table 1: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Week 24 to 31 March 3 to 9 April 10 to 16 April 17 to 23 April 24 to 30 April 1 to 7 May 8 to 14 May 15 to 21 May 22 to 28 May 29 May to 4 June 5 to 11 June 12 to 18 June 19 to 25 June 26 June to 02 July 03 to 09 July 10 to 16 July 17 to 23 July 24 to 30 July 31 July to 6 August 7 to 13 August Average Total Distribution of Responses by Week Total Number of Curriculums Total Number General Ads Sent of Calls Response Rate 56 448 60 13.39% 63 504 71 14.09% 65 520 32 6.15% 61 488 60 12.30% 61 488 92 18.85% 67 536 132 24.63% 73 584 116 19.86% 72 576 75 13.02% 74 592 98 16.55% 74 592 83 14.02% 72 576 135 23.44% 78 624 87 13.94% 73 584 90 15.41% 76 608 77 12.66% 73 584 63 10.79% 69 552 84 15.22% 68 544 101 18.57% 75 600 93 15.50% 66 528 45 8.52% 61 488 30 6.15% 69 551 81 14.65% 1377 11016 1624 15 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E The response rate varies from week to week, probably depending on the labor market cycles and economic expectations during the year. There is evidence that this expectations has been declining from March until now. Curriculums were sent to different job categories: professionals, technicians and unskilled workers. In the appendix we present a list of type of qualification within the different job categories. The average response rate by type of employment in Table 2 shows the same evolution as the response rate. We can also note that unskilled and technicians have a higher response rate than professionals. Professionals show a response rate of 12.1% compared to 14.2% for unskilled job announcements and 18.1% for technicians. Table 2: Number of CVs sent, Number of calls and Response rate by week and type of employment 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Week 24 to 31 March 3 to 9 April 10 to 16 April 17 to 23 April 24 to 30 April 1 to 7 May 8 to 14 May 15 to 21 May 22 to 28 May 29 May to 4 June 5 to 11 June 12 to 18 June 19 to 25 June 26 June to 02 July 03 to 09 July 10 to 16 July 17 to 23 July 24 to 30 July 31 July to 6 August 7 to 13 August Total Number of Curriculums Sent Number of Calls received Response Rate Professionals Technicians Unskilled Professionals Technicians Unskilled Professionals Technicians Unskilled 120 136 192 8 11 41 6.7% 8.1% 21.4% 176 168 160 7 18 46 4.0% 10.7% 28.8% 184 176 160 8 14 10 4.3% 8.0% 6.3% 168 160 160 2 21 37 1.2% 13.1% 23.1% 168 160 160 27 24 41 16.1% 15.0% 25.6% 200 176 160 34 63 35 17.0% 35.8% 21.9% 208 192 184 34 45 37 16.3% 23.4% 20.1% 192 200 184 22 32 21 11.5% 16.0% 11.4% 208 200 184 43 36 19 20.7% 18.0% 10.3% 192 200 200 15 52 16 7.8% 26.0% 8.0% 176 192 208 64 34 37 36.4% 17.7% 17.8% 208 200 216 24 51 12 11.5% 25.5% 5.6% 192 192 200 19 43 28 9.9% 22.4% 14.0% 216 192 200 35 34 8 16.2% 17.7% 4.0% 200 184 200 37 9 17 18.5% 4.9% 8.5% 168 184 200 23 39 22 13.7% 21.2% 11.0% 176 184 184 26 35 40 14.8% 19.0% 21.7% 208 192 200 19 52 22 9.1% 27.1% 11.0% 192 136 200 5 16 24 2.6% 11.8% 12.0% 176 112 200 11 19 0.0% 9.8% 9.5% 3728 3536 3752 452 640 532 12.1% 18.1% 14.2% Calls were received after different number of days. However, the following graph shows that more than 60% were received before day 10. This graph was done based on a Table shown in the Appendix. The average number of days in answering was 12 days approximately for professionals and unskilled was 14 days and for technicians was 8 days (see Table 3). 16 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Distribution of Days of Calls Unskilled .0 8 Professional .06 Frecuency .04 .02 0 0 50 100 150 Technicians .08 .06 .04 .02 0 0 50 100 150 Day Table 3 Number of Days they lasted in calling back Type of job Days Professionals Technicians Unskilled Average Day 14,02 8,69 14,81 Total Calls Back 452 640 532 Total CVs Sent 3728 3536 3752 Response Rate 12,12% 18,10% 14,18% Total 12,18 1624 11016 14,74% The resumes were sent by physical mail, email and fax. Table 4 shows the average number of days they lasted in making a call back by type of sending. We can see that CVs that were sent by physical mail received a call back in 18 days approximately and the ones sent by email received a call back in 8 days. The response to the fictitious candidates could have been done by telephone or by email. 17 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 4 Number of Days they lasted in calling back Way of sending them Days Phisical Mail Email Fax Average 18,70 8,12 17,00 Total Calls Back 621 1001 2 Total CVs Sent 3941 7059 16 Response Rate 15,76% 14,18% 12,50% Total 12,18 1624 11016 14,74% We will look at the average response rate by the three dimensions considered in this paper. 4.1 Gender Effects If we take a look at the gender based information, the response rates show very similar levels: 14.9% for men and 14.6% for women. This difference is small and not statistically significant (applying a test where the null hypothesis is the equality of the two proportions). In other words, men and women seem to have the same probability of getting called to interview. If the gender based difference is looked at in the response rates within the High Class group (by surnames), an slightly higher rate is obtained for women (15.3% vs 15.1%), in contrast to the situation with surnames in the Lower Classes were the response rates is higher for men (14.7% vs 14%). However, the differences are not statistically significant. 18 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 5 Callbacks by Gender Men Women Differences Diff Calls Diff Rate Z P-value -0.3% -0.6% 2.0% -2.0% 0.376 0.602 -1.572 1.778 0.707 0.547 0.116 0.075 5 5 15 -15 0.2% 0.5% 1.7% -1.6% -0.188 -0.349 -0.930 0.968 0.851 0.727 0.352 0.333 14.0% 10.7% 19.5% 12.0% -19 -17 21 -23 -0.7% -1.8% 2.4% -2.5% 0.731 1.228 -1.292 1.565 0.465 0.219 0.196 0.118 410 116 167 127 14.9% 12.4% 18.9% 13.5% -11 0 8 -19 -0.4% 0.0% 0.9% -2.0% 0.414 0.000 -0.491 1.244 0.679 1.000 0.623 0.213 395 104 171 120 14.3% 11.2% 19.3% 12.8% -3 -12 28 -19 -0.1% -1.3% 3.2% -2.0% 0.115 0.861 -1.742 1.272 0.908 0.389 0.082 0.203 CVs Sent Calls Rate Calls Rate General All Professionals Technicians Unskilled 5508 1864 1768 1876 819 232 302 285 14.9% 12.4% 17.1% 15.2% 805 220 338 247 14.6% 11.8% 19.1% 13.2% -14 -12 36 -38 High Social Class All Professionals Technicians Unskilled 2754 932 884 938 415 115 151 149 15.1% 12.3% 17.1% 15.9% 420 120 166 134 15.3% 12.9% 18.8% 14.3% Low Social Class All Professionals Technicians Unskilled 2754 932 884 938 404 117 151 136 14.7% 12.6% 17.1% 14.5% 385 100 172 113 High Income Mun. All Professionals Technicians Unskilled 2754 932 884 938 421 116 159 146 15.3% 12.4% 18.0% 15.6% Low Income Mun. All Professionals Technicians Unskilled 2754 932 884 938 398 116 143 139 14.5% 12.4% 16.2% 14.8% 4.2 Neighborhood Effects If we turn to the municipal dimension, the response rate of applicants from high income municipalities is 15.1%, compared to a rate of 14.4% for applicants from low income municipalities. These differences between municipalities, both on a general and on a cell level, are on average higher than that observed in the case of gender. However, this difference is not statistically significant to 90%. 19 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 6 Callbacks by Municipality CVs Sent High Income Municipality Calls Rate Low Income Municipality Calls Rate Differences Diff Calls Diff Rate Z P-value -0.7% -0.6% -0.7% -0.7% 1.021 0.602 0.524 0.655 0.307 0.547 0.600 0.512 -29 1 -9 -17 -1.1% 0.1% -1.0% -1.8% 1.092 -0.070 0.558 1.097 0.275 0.944 0.577 0.273 14.1% 10.9% 18.1% 13.4% -17 -13 -3 3 -0.6% -1.4% -0.3% 0.3% 0.652 0.939 0.185 -0.204 0.514 0.348 0.853 0.838 398 116 167 127 14.5% 12.4% 18.9% 13.5% -23 0 8 -19 -0.8% 0.0% 0.9% -2.0% 0.871 0.000 -0.491 1.244 0.384 1.000 0.623 0.213 395 104 171 120 14.3% 11.2% 19.3% 12.8% -15 -12 28 -19 -0.5% -1.3% 3.2% -2.0% 0.572 0.861 -1.742 1.272 0.567 0.389 0.082 0.203 General All Professionals Technicians Unskilled 5508 1864 1768 1876 831 232 326 273 15.1% 12.4% 18.4% 14.6% 793 220 314 259 14.4% 11.8% 17.8% 13.8% -38 -12 -12 -14 High Social Class All Professionals Technicians Unskilled 2754 932 884 938 430 117 163 150 15.6% 12.6% 18.4% 16.0% 401 118 154 133 14.6% 12.7% 17.4% 14.2% Low Social Class All Professionals Technicians Unskilled 2754 932 884 938 405 115 163 123 14.7% 12.3% 18.4% 13.1% 388 102 160 126 Men All Professionals Technicians Unskilled 2754 932 884 938 421 116 159 146 15.3% 12.4% 18.0% 15.6% Women All Professionals Technicians Unskilled 2754 932 884 938 410 116 143 139 14.9% 12.4% 16.2% 14.8% 4.3 Social Class Effect A similar situation to the above may be observed when the response rates of fictitious candidates with Upper Class surnames (15.2%) are compared with those with Lower Class surnames (14.3%). Once again, the differences are not statistically significant. The largest differences occur within the group of women and also within the high income municipalities category. A similar situation to the above may be observed when the response rates of fictitious candidates with Upper Class surnames (15.2%) are compared with those with Lower Class surnames (14.3%). Once again, the differences are not statistically significant. The largest differences occur within the group of women and also within the high income municipalities category. 20 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 7 Callbacks by Surname CVs Sent High Social Class Calls Rate Low Social Class Calls Rate Differences Diff Calls Diff Rate Test Z P-value General All Professionals Technicians Unskilled 5508 1864 1768 1876 835 235 317 283 15.2% 12.6% 17.9% 15.1% 789 217 323 249 14.3% 11.6% 18.3% 13.3% -46 -18 6 -34 -0.8% -1.0% 0.3% -1.8% 1.236 0.903 -0.262 1.591 0.216 0.367 0.793 0.112 High Income Mun. All Professionals Technicians Unskilled 2754 932 884 938 430 117 163 150 15.6% 12.6% 18.4% 16.0% 405 118 154 133 14.7% 12.7% 17.4% 14.2% -25 1 -9 -17 -0.9% 0.1% -1.0% -1.8% 0.939 -0.070 0.558 1.097 0.348 0.944 0.577 0.273 Low Income Mun. All Professionals Technicians Unskilled 2754 932 884 938 401 115 163 123 14.6% 12.3% 18.4% 13.1% 388 102 160 126 14.1% 10.9% 18.1% 13.4% -13 -13 -3 3 -0.5% -1.4% -0.3% 0.3% 0.500 0.939 0.185 -0.204 0.617 0.348 0.853 0.838 Men All Professionals Technicians Unskilled 2754 932 884 938 415 115 151 149 15.1% 12.3% 17.1% 15.9% 404 117 151 136 14.7% 12.6% 17.1% 14.5% -11 2 0 -13 -0.4% 0.2% 0.0% -1.4% 0.417 -0.140 0.000 0.836 0.677 0.889 1.000 0.403 Women All Professionals Technicians Unskilled 2754 932 884 938 420 120 166 134 15.3% 12.9% 18.8% 14.3% 385 100 172 113 14.0% 10.7% 19.5% 12.0% -35 -20 6 -21 -1.3% -2.1% 0.7% -2.2% 1.335 1.436 -0.363 1.434 0.182 0.151 0.717 0.152 In conclusion, surprisingly, relevant gender differences are not found. In addition, the differences in response rates by municipalities or surnames are lower than the gender differences, in fact they are not statistically significant. The analysis of the response rates for professionals confirms in general the aspects found above. There are no significant differences by gender, municipality or by surname. 4.4 Regression Analysis Table 8 runs some complementary analysis using regression. As it can be seen, in none of the specifications there is a change in the main conclusions. The dummy variables associated to gender, municipality or surname are not significant. 21 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 8 Regressions for the probability of receiving a callback (Dependent Variable: Dummy=1 if a callback is received) Variable Coeff. p-value Coeff. p-value Coeff. p-value Coeff. p-value Dummy High Income Municipality=1 Dummy Men=1 Dummy High Class Surname=1 Dummy Professional Job ad=1 Dummy Technician Job ad=1 Dummy Studied at Private School=1 Dummy Studied at Municipal School=1 Controls for type of mail sent Including interactions Pseudo R2 Number of observations 0.0069 0.0026 0.0082 No No 0.0003 11016 0.304 0.706 0.222 0.0070 0.0029 0.0084 -0.0217 0.0380 No No 0.006 11016 0.301 0.670 0.210 0.009 0.000 0.0074 0.0020 0.0082 -0.0262 0.0369 -0.0030 -0.0173 Yes No 0.0189 11016 0.392 0.0048 0.736 0.770 -0.0106 0.389 0.226 -0.002 0.863 0.003 -0.0249 0.004 0.000 0.0370 0.000 0.780 -0.0114 0.741 0.038 0.0061 0.668 Yes Yes 0.009 11016 Note: Probit Regressions. Coefficients are expressed in probability points for discrete changes of dummy variables from 0 to 1 (evaluated at means). 4.5 Timing of the callbacks The results shown until now allow us to say that there are no differences in callback rates across groups. However, it could be possible to hypothesize differences favoring some groups in the timing of the callbacks. This is not the case, however, as it is shown in.Table 9. All the differences reported in the number of days to receive a callback across groups are not statistically significant. 22 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 9 Number of days to receive a callback Mean Median Gender: Men Women Difference 12.8 11.6 1.2 8 7 1 Municipality: High Income Low Income Difference 11.8 12.5 -0.7 7 7 0 Surname: High Class Low Class Difference 12.3 12.1 0.2 7 7 0 23 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E References Adimark, “Mapa Socioeconómico de Chile. Nivel socioeconómico de los hogares del país basado en datos del Censo”. Anderson, Lisa; Roland Fryer and Charles Holt (2005). “Discrimination: Experimental Evidence from Psychology and Economics.” Forthcoming in Handbook on Economics and Discrimination, William Rogers, Ed. Antonovics, Kate; Peter Arcidiacono and Randy Walsh (2004). “Competing Against the Opposite Sex.” Economics Working Paper Series 2003-08, University of California at San Diego. Antonovics, Kate; Peter Arcidiacono and Randy Walsh (2005). “Games and Discrimination: Lessons from the Weakest Link.” Forthcoming at Journal of Human Resources. Altonji, Joseph and Rebecca Blank (1999). “Race and Gender in the Labor Market.” Handbook of Labor Economics, 3, pp. 3143-3259. Arenas, Alberto, Jere Behrman and David Bravo (2004) “Characteristics of and Determinants of the Density of Contributions in a Private Social Security System”. Working Paper, Michigan Retirement, Research Center, May, 2004. Becker, Gary (1971) The Economics of Discrimination, 2nd Edition, The University of Chicago Press, IL. Becker, Gary (1991) A Treatise on the Family, enlarged edition: Harvard University Press, Cambridge, MA. Bertrand, Marianne and Sendhil Mullainathan (2004). “Are Emily and Greg more Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” American Economic Review, 94(4), pp. 991-1013(23). Blank, Rebecca; Marilyn Dabady and Constance Citro, Eds. (2004). “Measuring Racial Discrimination. Panel on Methods for Assessing Discrimination.” The National Academies Press, Washington, D.C. Blinder, Alan (1973). “Wage Discrimination: Reduced Form and Structural Estimates.” The Journal of Human Resources, 7(4), pp. 436-55. 24 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Bravo, David (2005) “Elaboración, Validación y Difusión de Índice Nacional de Calidad del Empleo Femenino”, Centro de Microdatos, Universidad de Chile. Report prepared to the Secretary of Gender Sigues (Ministerio Servicio Nacional de la Mujer). Bravo, David (2004), “Análisis y principales resultados. Primera Encuesta de Protecciòn Social”. Departamento de Economía, Universidad de Chile y Ministerio del Trabajo y Previsiòn Social, Julio. Contreras, Dante y Gonzalo Plaza (2004) “Participación Femenina en el Mercado Laboral Chileno. ¿Cuánto importan los factores culturales?”, Departamento de Economía, Universidad de Chile Contreras, Dante y Esteban Puentes (2001) “Is The Gender Wage Discrimination Decreasing In Chile? Thirty Years Of ‘Robust’ Evidence”, Departamento de Economía, Universidad de Chile. Fernandez, Fogli And Olivetti (2004) “Preference Formation And The Rise Of Women’s Labor Force Participation: Evidence From WWII”, NBER Working Paper 10589 Goldin, Claudia and Cecilia Rouse (2000). “Orchestrating Impartiality: The Impact of ‘Blind’ Auditions on Female Musicians.” American Economic Review 90(4), pp. 715-741. Heckman, James, and Peter Siegelman (1993). “The Urban Institute Audit Studies: Their Methods and Findings.” In Clear and Convincing Evidence: Measure of Discrimination in America. Michael Fix and Raymond Struyk, editors. The Urban Institute Press, Washington D.C. Heckman, James (1998). “Detecting Discrimination.” The Journal of Economic Perspectives,12(2), pp. 101-116. Heckman, James, Jora Stixrud and Sergio Urzua (2005) “The Effects of Cognitive and Noncognitive Abilities on Labor Market Outcomes and Social Behavior”, University of Chicago. Levitt, Steven (2004). “Testing Theories of Discrimination. Evidence from ‘The Weakest Link.’” Journal of Law and Economics, 47, pp. 431-452. List, John (2003). “The Nature and Extent of Discrimination in the Marketplace: Evidence from the Field.” Quarterly Journal of Economics, 119(1), pp. 49-89. 25 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Moreno, Martin; Hugo Ñopo, Jaime Saavedra and Maximo Torero (2004) “Gender and Racial Discrimination in Hiring. A Pseudo-Audit Study for Three Selected Occupations in Metropolitan Lima.” IZA Discussion Paper 979. Montenegro, Claudio (1999) “Wage distribution in Chile: Does Gender Matter? A Quantile Regression Approach. Mimeo, Universidad de Chile. Montenegro, Claudio y Paredes, Ricardo (1999) “Gender Wage Gap and Discrimination: A Long Term View Using Quantile Regression”. Mimeo, Universidad de Chile. Neal, Derek A. and William R. and Johnson (1996) “The Role of Premarket Factors in Black-White Wage Differences”, The Journal of Political Economy, Vol. 104, No. 5 (Oct., 1996), 869-895. Newmark, David; Roy J. Bank and Kyle D. Van Nort (1996) “Sex Discrimination in Restaurant Hiring: An Audit Study”, The Quarterly Journal of Economics, Vol. 111, No. 3 (Aug., 1996), 915-941. Nuñez, Javier and Roberto Gutierrez (2004) “Classism, Discrimination and Meritocracy in the Labor Market: The Case of Chile.” Documento de trabajo 208. Departamento de Economia, Universidad de Chile. Ñopo, Hugo (2004). “Matching as a Tool to Decompose Wage Gaps.” IZA Discussion Paper No. 981. Oaxaca, Ronald (1973). “Male-Female Wage Differentials in Urban Labor Market.” International Economic Review, 14(3), pp. 693-709. O’Neil, June E. and Dave M. O’Neil (2005) “What do wage Differentials tell us about labor Market Discrimination”, NBER Working Paper 11240. Paredes, Ricardo y Riveros, Luis (1994): “Gender Wage Gaps in Chile. A Long term View:1958:1990”. Estudios de Economía, Vol.21, Número especial, 1994. Riach, Peter and Judith Rich (2002). “Field Experiments of Discrimination in the Marketplace.” The Economic Journal,112, pp. 480-518. Riach, Peter and Judith Rich (2004). “Deceptive Field Experiments of Discrimination: Are they Ethical?” KYKLOS, 57(3), pp. 457-470. Rosenberg, M. (1965). Society and the Adolescent Self-Image. Princeton, NJ: Princeton University Press. 26 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Rotter, J. B. (1966). Generalized Expectancies for Internal versus External Control of Reinforcement. Washington DC: American Psychological Association. 27 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Appendix to Chapter 1 Table A.1 Unskilled Number Administrativo Aseador Auxiliar Aseo Bodeguero Cajero Cobrador Conductor Conductores Digitador Encuestador Fotocopiador Garzon Garzón Guardia Operario Producción Operario Tintoreria Promotor Recepcionista Recepcionistas Vendedor Volantero Total 952 208 48 384 328 96 32 16 368 88 8 112 40 56 8 8 304 8 8 624 56 3,752 % 25.37 5.54 1.28 10.23 8.74 2.56 0.85 0.43 9.81 2.35 0.21 2.99 1.07 1.49 0.21 0.21 8.1 0.21 0.21 16.63 1.49 100 28 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table A2 Professionals Number Abogado Abogado litigante Abogado media Jornada Abogado part-time Constructor Civil Constructor Civil (jefe proyecto) Constructor Civil de Obra Constructor Civil en altura Contador Auditor Contador Auditor Bilingüe Ing. Civil Electronico Ing. Civil Informatico Ing. Civil Informático Ing. Civil Telecomunicaciones Ing. Comercial (Marketing) Ing. Ejec. En Computacion Ing.Comercial Marketing Ingeniero Civil Ingeniero Civil Computacion Ingeniero Civil Constructor Ingeniero Civil Industrial Ingeniero Civil en Computacion Ingeniero Comercial Ingeniero Comercial MBA Ingeniero Constructor Ingeniero Ejec Informatico Ingeniero Ejec. Informatico Ingeniero Ejec. Informático Ingeniero Electronico Ingeniero Informatico Ingeniero Informático Ingeniero Informático (Teradata) Ingeniero Obras Civiles Ingeniero Telecomunicaciones Ingeniero en Computacion Ingeniero en Telecomunicaciones Ingeniero, Const. Civil Profesor Psicologo Psicólogo Supervisor Educacional Total % 168 8 8 8 600 8 8 8 905 7 8 32 48 8 8 8 8 104 16 8 24 8 552 8 8 16 24 72 8 136 104 8 8 8 8 8 8 720 8 8 8 3728 4.51 0.21 0.21 0.21 16.09 0.21 0.21 0.21 24.28 0.19 0.21 0.86 1.29 0.21 0.21 0.21 0.21 2.79 0.43 0.21 0.64 0.21 14.81 0.21 0.21 0.43 0.64 1.93 0.21 3.65 2.79 0.21 0.21 0.21 0.21 0.21 0.21 19.17 0.21 0.21 0.21 100 29 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table A3 Technicians Number Soporte Computacional 8 Administrador 16 Administrador Empresas 8 Administrador Sistema 8 Administrador de Botilleria 8 Administrador de Empresas 8 Administrador de Local 16 Administrador de Redes 16 Administrador de Restaurant 8 Administrador de Sistemas 16 Administrador de red 8 Administrador de redes 8 Administrativo en Comex 8 Adquisiciones 8 Agente de Ventas 16 Agente de Ventas Intangibles 8 Analista Computacional 8 Analista Programador 200 Analista Sistemas 8 Analista de Sistema 32 Analista de Sistemas 24 Analista o Programador 8 Asesor Comercial Marketing 8 Asistente Adquisiciones 16 Asistente Comercio Exterior 8 Asistente Contable 40 Asistente Técnico Hardware 8 Asistente de Enfermeria 8 Asistente de Enfermos 16 Auxiliar Enfermería 8 Auxiliar Paramedico 16 Auxiliar Paramédico 32 Auxiliar Técnico de Laboratorio 8 Auxiliar de Enfermeria 40 Auxiliar de Enfermería 40 Auxiliar de Laboratorio 8 Auxiliar de enfermería 8 Auxiliar de laboratorio 8 Auxiliar de toma de muestra 8 Ayudante Contable 8 Ayudante de Contador 40 Chef 32 Cheff Ejecutivo 8 Comercio Exterior 8 Conocimientos en Computacion 8 Contador 200 Contador Administrador 8 Contador Asistente 16 Contador General 72 Contador general 8 Desarrollador de Web 8 Dibujante Autocad 48 Dibujante Estructural 8 Dibujante Gráfico 8 Dibujante Mecánico Autocad 8 Dibujante Proyecticta 8 Dibujante Técnico 32 Dibujante de Arquitectura 8 Dibujante técnico 24 Dibujante y Proyectistas 8 Diseñador Gráfico 128 Diseñador Industrial 32 Diseñador Internet 8 Diseñador Web 16 Diseñador Web Master 8 Diseñador de Página web 8 Diseñador de web 8 Ejecutivo Comercio Exterior 8 Ejecutivo Telemarketing 8 Ejecutivo de Ventas 8 Encargado de Adquisiciones 16 Encargado de Adquisisciones 8 Encargado de Compras 8 Encargado de Informatica 8 Encargado de Informática 8 Encargado de Local 8 Encargado de Remuneraciones 8 Encargado de comercio exterior 8 Encargado de informática 8 Encargado de remuneraciones 8 Experto en Computación 8 Experto en Diseño Página Web 8 Explotador de Sistemas 8 Informático 8 % 0.23 0.45 0.23 0.23 0.23 0.23 0.45 0.45 0.23 0.45 0.23 0.23 0.23 0.23 0.45 0.23 0.23 5.66 0.23 0.9 0.68 0.23 0.23 0.45 0.23 1.13 0.23 0.23 0.45 0.23 0.45 0.9 0.23 1.13 1.13 0.23 0.23 0.23 0.23 0.23 1.13 0.9 0.23 0.23 0.23 5.66 0.23 0.45 2.04 0.23 0.23 1.36 0.23 0.23 0.23 0.23 0.9 0.23 0.68 0.23 3.62 0.9 0.23 0.45 0.23 0.23 0.23 0.23 0.23 0.23 0.45 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 0.23 Informático Hardware Jefe Adquisiciones Jefe Facturación Jefe de Abastecimiento Jefe de Bodega Jefe de Local Jefe de Locales Jefe de Personal Jefe de Recursos Humanos Jefe de Tienda Jefe de Tiendas Jefe para cafeteria y pasteleria Operador Informático Paramedico Paramedico RX Paramedicos Pedidor Aduanero Prevencionista Riesgos Procurador Programador Programador Analista Programador Clipper Programador Web Programador Webmaster Programador o Analista Programador y Analistas Proyectista Autocard Soporte Soporte Computacional Soporte Informático Soporte Tecnico Soporte Técnico Soporte en Redes Supervisor Supervisor Cobranzas Supervisor Locales Comerciales Supervisor Logístico Supervisor de Call Center Supervisor de Facturación y cobranzas Supervisor de Venta Tecnico Informatico Tecnico Paramedico Tecnico Paramedicos Tecnico Soporte Tecnico en Computación Tecnico en Redes Tecnico paramedico Técnico Administración de Redes Técnico Administrador Empresas Técnico Comercio Exterior Técnico Computación Técnico Gastronómico Técnico Informático Técnico Instalación Redes Técnico Jurídico Técnico Paramédico Técnico Prevención Técnico Programador Técnico Químico Técnico Soporte Terreno Técnico Soporte en Linux Técnico de Comercio Exterior Técnico en Comercio Exterior Técnico en Comex Técnico en Computación Técnico en Computación y Redes Técnico en Enfermería Técnico en Gastronomía Técnico en Hardware y Redes Técnico en Hardware y Software Técnico en Informática Técnico en Logística Técnico en Mantención Técnico en Programación Técnico en Redes Computacionales Técnico en Reparación Técnico en Soporte Técnico en Soporte Computacional Técnico en comex Técnico paramédico Técnico pc grafico Vendedores Isapre Web Master Total Number 8 8 8 8 8 56 8 8 8 32 8 8 8 16 8 8 8 8 32 544 8 8 80 8 8 8 8 16 88 8 8 24 16 8 24 8 16 8 8 8 8 8 8 16 8 8 8 8 8 32 16 8 32 8 24 88 8 24 8 8 8 8 16 8 128 8 8 8 8 8 16 8 8 8 8 8 72 8 8 8 8 8 8 1648 % 0.23 0.23 0.23 0.23 0.23 1.58 0.23 0.23 0.23 0.9 0.23 0.23 0.23 0.45 0.23 0.23 0.23 0.23 0.9 15.38 0.23 0.23 2.26 0.23 0.23 0.23 0.23 0.45 2.49 0.23 0.23 0.68 0.45 0.23 0.68 0.23 0.45 0.23 0.23 0.23 0.23 0.23 0.23 0.45 0.23 0.23 0.23 0.23 0.23 0.9 0.45 0.23 0.9 0.23 0.68 2.49 0.23 0.68 0.23 0.23 0.23 0.23 0.45 0.23 3.62 0.23 0.23 0.23 0.23 0.23 0.45 0.23 0.23 0.23 0.23 0.23 2.04 0.23 0.23 0.23 0.23 0.23 0.23 100 30 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table A4 Days Number of Days they lasted in calling back Type of job Professionals Technicians Unskilled 0 10 90 54 1 55 92 65 2 11 57 45 3 10 36 44 4 3 19 15 5 14 20 7 6 26 22 15 7 19 58 50 8 31 50 21 9 26 23 28 10 17 11 22 11 31 5 4 12 7 5 2 13 11 5 5 14 24 28 15 15 9 24 11 16 12 13 11 17 9 7 5 18 11 3 19 2 2 1 20 15 2 21 7 4 12 22 13 4 7 23 5 1 3 24 9 5 26 1 9 1 27 18 4 3 28 3 6 5 29 1 3 2 30 9 1 5 31 1 32 1 33 1 9 34 4 Total 154 212 113 90 37 41 63 127 102 77 50 40 14 21 67 44 36 21 14 5 17 23 24 9 14 11 25 14 6 15 1 1 10 4 31 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 35 36 37 38 40 41 42 43 44 48 49 50 51 52 54 55 57 58 59 64 66 73 74 76 77 84 85 86 90 91 93 95 98 105 111 116 125 126 Average Day Total Calls Back Total CVs Sent Response Rate 2 7 2 3 2 2 1 2 1 1 1 5 1 5 1 4 2 2 1 3 1 1 2 4 1 4 1 3 1 3 5 3 14,02 452 3728 12,12% 8,69 640 3536 18,10% 1 4 8 2 4 4 1 2 1 2 1 2 4 1 1 1 2 1 1 1 1 14,81 532 3752 14,18% 3 9 7 3 3 7 2 5 2 4 2 5 4 1 4 1 4 9 5 1 3 4 4 1 7 4 2 1 2 4 1 1 1 2 1 1 1 1 12,18 1624 11016 14,74% 32 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table A5 Days Number of Days they lasted in calling back Way of sending them Phisical Mail Email Fax 0 154 1 212 2 4 109 3 47 43 4 26 11 5 16 25 6 19 44 7 66 61 8 54 48 9 61 16 10 21 29 11 19 21 12 2 12 13 4 17 14 27 40 15 29 15 16 20 16 17 9 10 2 18 10 4 19 3 2 20 10 7 21 11 12 22 17 7 23 5 4 24 11 3 26 9 2 27 6 19 28 8 6 29 5 1 30 14 1 31 1 32 1 33 1 9 34 4 35 3 36 4 5 37 7 154 212 113 90 37 41 63 127 102 77 50 40 14 21 67 44 36 21 14 5 17 23 24 9 14 11 25 14 6 15 1 1 10 4 3 9 7 33 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 38 40 41 42 43 44 48 49 50 51 52 54 55 57 58 59 64 66 73 74 76 77 84 85 86 90 91 93 95 98 105 111 116 125 126 Average Total Calls Back Total CVs Sent Response Rate 2 6 2 1 3 1 5 2 2 2 2 4 1 2 3 4 1 4 8 3 4 4 1 7 3 2 1 2 4 1 1 1 2 1 3 1 1 2 1 1 1 18,70 621 3941 15,76% 1 8,12 1001 7059 14,18% 17,00 2 16 12,50% 3 3 7 2 5 2 4 2 5 4 1 4 1 4 9 5 1 3 4 4 1 7 4 2 1 2 4 1 1 1 2 1 1 1 1 12,18 1624 11016 14,74% 34 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Chapter 2 “Is there labor market discrimination among professionals in Chile? Lawyers, Doctors and Business-people” Abstract This paper presents a complete analysis of the gender differences in three Chilean professionals labor market: Business, Law and Medicine. In the analysis, we utilize a new and rich data set collected for this effects. This data set contains information on labor market outcomes (including labor history), on schooling attainment and schooling performance, on a complete set of variables characterizing the family background of the individuals in the sample and on non cognitive abilities. Our results show that differences in wages attributed to gender are only present in the Law profession. In the Business/Economics profession a vector of current family condition makes the gender effect disappear and in Medicine taking into account hours worked, size of the firm and region make also disappear the gender gap. Specially important are shown to have a better level of self control in explaining wage differences. 35 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 1 Introduction Labor market discrimination is said to arise when two identically productive workers are treated differently on the grounds of the worker’s race or gender, when race or gender do not in themselves have an effect on productivity (Altonji and Blank, 1999; Heckman, 1998). However, there are never identical individuals. There are several unobservable factors that determine individual performance in the labor market. First, we do not observe individual’s cognitive abilities 14 . Second, we do not observe individual’s non-cognitive abilities such as personal motivation, self-determination, and locus of internal/external control or selfconfidence. Third, we do not observe pre-labor market discrimination conditions such as family background and school environment 15 . Fourth, we do not observe individual past expectations about how the labor market works 16 . Regarding gender group differences, these can be found for market and non-market activities and for types of jobs. There are gender differences for comparative advantages due to: differences in gender roles in home production, differences in parental investment in skills (Becker, 1991) and the transfer of family preferences (Fernandez, Fogli and Olivetti, 2004). And there are group gender differences in human capital investments as a result of pre-labor market discrimination. Consequently, discrimination can influence human capital investment before and after an individual enters the labor market. Based on these facts and on the lack of studies in Chile which can face these issues, we implemented a survey on professionals from three different careers in Chile: Law, Medicine and Business, to analyze differences in their wages but reducing unobservable to a minimum. They have all passed a university entrance selection test. They are comparable in their academic formation. We have data on their university performance. We have data on their social and family background. We have applied tem a test on non-cognitive abilities. We have applied a survey to ask them real labor experience and family conditions now. 14 Neal and Johnson (1996) is a good example of how unobserved factors could be driving the results. They study the role of pre-market factors in black-white wage differences controlling with a test administered to teenagers prepared to leave high school in the US. They found that the adult black-white wage gap primarily reflects a skills gap due to observable differences in family backgrounds and school environments. 15 O’Neil and O’Neil (2005) find that differences in productivity-related factors account for most of the between-group wage differences in the year 2000 for the US. Differences in schooling and in skills developed in the home and in school, as measured by test scores, are important in explaining black/white wage gaps. But the gender differences in schooling and cognitive skills are quite small and explain little of the pay gap. Instead the gender gap is largely due to choices made by women and men concerning the amount of time and energy devoted to a career, as reflected in years of work experience, use of part-time work, and other workplace and job characteristics. 16 See Altonji and Blank (1999) for a complete survey on race and gender discrimination and explanations of the underlying theories. 36 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Based on this complete and new dataset we have taken a regression analysis approach to determine how much of the wage gap is left once the only difference among individuals is gender. Research in Chile have been centered on the traditional Oaxaca decomposition (Oaxaca, 1973; Blinder, 1973). Paredes and Riveros (1993), estimate the endowment and discrimination effects for the period 1958-1990 17 . They provide evidence on discrimination against females during the whole period examined. Montenegro (1999) and Montenegro and Paredes (1999) analyze the gender wage differential by using quantile regression and the Oaxaca decomposition. The evidence also shows stable and systematic differences in the returns to education and to experience by gender along the conditional wage distribution. In addition, it is also shown that discrimination is higher for women with more education and experience. However, these conclusions of studies are limited. They lack several control variables, related to cognitive and non-cognitive abilities and school and family environments. In addition, preferences over non-market activities and experience of Chilean female workers could prove to be a very important unobservable factor. More recently, Núñez and Gutiérrez (2004) study social class discrimination in Chile under the traditional Blinder-Oaxaca decomposition. They use a dataset that allows them to reduce the role of unobservable factors by limiting the population under study and having better measures of productivity as we do. However, this study has some limitations. One it is related to the collection of the data. The survey was carried out by postal mail and had a very low response rate, 30% approximately. Second, the survey was carried out on recently graduated college students of Economics 18 which does not allowed to detect the effects of labor experience. Third, it lacks of survey data on labor history and real experience, family characteristics and preferences. Fourth, the survey had a very small sample size. This paper faces these limitations by surveying 1,500 Alumni of the Universidad de Chile from the following degree programs: 500 from Medicine, 500 from Law and 500 from Business/Economics. Half of each group are women and half are men. We subsequently analyze wage differences between women and men for each careers correcting the estimates for post graduate schooling, labor market experience, parents schooling, married conditions and cognitive abilities. Following recent literature (Heckman, Stixrud and Urzua, 2005) we took the Rotter (1966) and Rosemberg (1965) tests for non-cognitive abilities. We have run OLS regressions and ordered probit estimation to explain economic outcomes by a set of explanatory variables. 17 18 Contreras and Puentes (2001) extended the analyses to 1996. In Chile, high school students choose subjects, not colleges as in the US. 37 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E The results indicate that that differences in wages attributed to gender are only present in the Law profession. In the Business/Economics profession a vector of current family condition makes the gender effect disappear and in Medicine taking into account hours worked, size of the firm and region make also disappear the gender gap. Specially important are shown to have a better level of self control in explaining wage differences. This structure of the paper is the following. Next section 2 present the econometric models. Section 3 presents the data and summary of the descriptive statistics. Section 4 presents the results and finally section 5 presents the conclusions. 2 The Econometric Models In this section we will explain briefly the well known models in which it is usually study labor market discrimination. We are using two different specifications: OLS estimation and an ordered probit estimation 19 . In each of these models we have a wage equation as a function of a set of different explanatory variables: Model 1: OLS log wi = γ Fi + λ1 Exp + λ2 Exp 2 + λ3 N i jobs + J 'i Φ + X i'δ + S 'i Γ + T 'i Δ + H 'i Π + ε i where F is a dummy variable that takes value 1 if female and 0 otherwise. Thus, the coefficient γ measured the perceptual difference in wages that is lower because individual i is female rather than male. In this setting, it is assumed that the market value in the same way the characteristics of the individuals. Exp is years of real labor experience and Exp2 is the squared. And Njobs is the average number of parallel working activities each individual does in a year. J’ is a set of variables related to characteristics of the job, it contains a dummy variable for the level of responsibility in the job which take value 1 if the occupation is of high responsibility 20 and 0 otherwise, a dummy variable for a big firm that takes value 1 if the firm 19 In a future version of this paper we will include a Oaxaca decomposition. An occupation is set to be of high responsibility if its occupation code is related to the following categories: members of the executive or legislative power and directives of public and private firms, such as managers of business and company directors. 20 38 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E has more than 500 workers and a dummy variable equals to 1 if the person works in the metropolitan region. X’ is a set of variables related to other personal characteristics such as a dummy variable that takes value 1 if the person has done a postgraduate course, university performance measured with a dummy variable that takes a value 1 if the person reproved a class and age. S’ is a set of variables related to the person socioeconomic background such as mother’s and father’s years of schooling and grades at secondary school 21 . T’ contains two measures of non cognitive abilities explained later. Finally, H’ contains three measures of current family situation such as a dummy for married, number of children and a dummy for head of the household. An alternative specification would have been to have a Heckman model Model 2: Ordered Probit Model I i = j if where α j +1 ≤ φ Hrsi + γ Fi + λ1 Exp + λ2 Exp 2 + λ3 N i jobs + J 'i Φ + X i'δ + S 'i Γ + T 'i Δ + H 'i Π + ε i < α j j = 1,K ,8 Ii is an indicator variable for the wage intervals and Hrsi is the monthly hours worked by individual i. 3 The Data In this section we present a comprehensive descriptive statistics of the variables collected in the survey and used in the estimations 22 . We will look at different statistics of labor market outcomes, performance at University, social and academic background, test for non cognitive abilities and current household status. Each of these variables is meant to explain in same way differences between observed gender gaps in wages. 21 22 Grades is Chile go from 1 to 7, having an average of secondary school performance of 6 is distinction. The questionnaire and a complete field work resume are in the appendix. 39 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E We have collected approximately the same quantity of interviews for each type of degree (see Table 1). That is 500 observations for each type. Table 1 All Sample Type of Degree Obs % Business 505 33.18 Law 506 33.25 Medicine 511 33.57 Total 1522 100 Table 2 shows the list of variables included in the regression for the degree of Business and Economics by gender. Regarding labor outcomes we can see that there are gender differences on wages 23 . Women’s monthly wage is 69% of men’s monthly wage, these differences can also be observed in the tabulations of wage intervals. However, since women work less hours a month, women’s hourly wage is only 81% of men’s hourly wage, and this is 97% if we look at the logarithm. We also can note that female labor force participation is 81% and is lower than male labor force participation which is 97%. Women have less accumulated experience and have less parallel activities, although these differences are not high. 56% of men have a job of high responsibility while 43% of women have the same level of responsibility in the job. We can also observe that there are differences in the type of firm they work. 47% of men work in firms with more than 500 workers, while 60% of women do the same. We can note that more 15% of less women do a post graduate degree, although women seems to have a better performance at University and at school (see grades). Mother’s schooling is higher for women than for men. These latter may be related to the transmission of preferences. There are not differences in socioeconomic background between men and women. 8% of each group comes from a poor family and 15% of each group was raised in a uni-parental home. As we said before, we also collected measures of non-cognitive abilities by taking the Rotter (1966) and Rosenberg (1965) tests for internal and external locus of control and self-esteem, respectively 24 . The lower the index the higher is the degree of self control or self esteem. We 23 Exact wages were only reported for 20% of the sample approximately, however most people who did not give the exact amount gave an interval. Therefore we have assigned the maximum of the interval to the wage. We are anyway running ordered probit using the intervals of wages. There were also people who did not want to answer this question therefore we have less data for this variable. 24 The tests are included in the questionnaire. 40 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E can note than in average women got a lower degree of self control but a higher degree of self esteem. Finally, we think that measures of the current home situation could be important since it may reflect preferences for home production activities. We can see that although the number of children and the percentage of married men and women are the same only 28% of women are head of the household whereas 96% of men are in the same situation. 41 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 2 Summary Statistics: Business/Economics Male Female Obs Mean SD Obs Mean SD Labor Market Outcomes Hourly Wage 211 12120.4 4760.4 182 9842.93 5695.5 Log(hourly wage) 211 9.33 0.40 182 9.07 0.53 Monthly Wage 211 2314882 905585 182 1602061 756934 Labor Market Participation 252 0.97 0.16 253 0.85 0.36 Monthly Hours worked 245 192.56 31.71 214 184.17 265.63 Real experience 252 17.50 5.31 253 16.96 4.54 Real experience squared 252 334.38 207.88 253 308.15 166.66 Mean of number jobs by year 252 1.05 0.34 253 1.00 0.32 Level of responsibility 245 0.56 0.50 214 0.43 0.50 Big Firm (>500w) 252 0.47 0.50 253 0.60 0.49 Metropolitan Region 252 0.92 0.27 253 0.93 0.25 Age 252 42.50 6.40 253 41.04 5.20 Performance at University Reprove any class==1 252 0.89 0.31 253 0.83 0.38 Post graduate schooling==1 252 0.47 0.50 253 0.32 0.47 Family Background Mother's years of schooling 248 12.95 3.20 249 13.34 3.25 Father's years of schooling 245 14.62 3.26 247 14.58 3.55 Grades in secondary school 245 60.21 4.06 252 63.64 2.74 Poor background==1 250 0.07 0.26 253 0.08 0.26 Uniparental home==1 252 0.16 0.37 253 0.15 0.35 Non Cognitive Abilities Self control test 247 1.34 0.41 248 1.42 0.43 Self esteem test 245 1.55 0.38 249 1.49 0.40 Family Status Number of children 247 2.26 1.59 246 2.28 1.39 Married==1 252 0.85 0.35 253 0.82 0.38 Head of the household==1 252 0.96 0.19 253 0.28 0.45 Wage Intervals (1USD=537CHP) % % Less than 372 USD 0.47 1.1 Between 372 and 745 USD 1 2 3.85 Between 745 and 1120 USD 7 0.47 4.95 Between 1120 and 1490 USD 1 9 2.37 10.99 Between 1490 and 1862 5 20 9 21.43 Between 1862 and 2793 USD 19 39 19.91 20.88 Between 2793 and 3725 USD 42 38 19.91 17.58 Between 3725 and 4656 USD 42 32 17.06 9.89 Between 4656 and 5587 USD 36 18 More thab 5587 USD 65 30.81 17 9.34 Total 211 100 182 100 42 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 3 shows the summary of the descriptive statistics for the degree of Law by gender. In this case the gap in monthly wages is 68% approximately in favour of men. However, we can note that monthly hours worked by women are in average higher than hours worked by men and so the gap reduces to 71% in monthly hourly wage and to 96% if we look at the logarithm. We also can note that female labor force participation is 93% and is lower than male labor force participation which is 99%, both are higher than in case of business. Women have more accumulated experience and have slightly less parallel activities, although these differences are also not high. We can also observe that the proportion of lawyers in job positions with more responsibility is less than in the case of business/economics reaching only 4% and 5% respectively. We can also observe that there are differences in the type of firm they work. In this case, women also tend to work in big firms (51%) more than men (31%). We can note that 63% of women and men that study law do post graduate degrees. Again women have a better performance at University and at school: a lower proportion of women reprove classes and they have higher grades at secondary school. Mother’s and father’s schooling are higher for women than for men. This may be again related to the transmission of preferences. Only 6% of women come from a family of poor background, whereas 17% of men are in the same situation. 20% of each group was raised in a uni-parental home. The measures of non cognitive abilities behave in the same way. In average women got a lower degree of self control but a higher degree of self esteem. Finally, regarding the measures of current home situation present the following characteristics. We can see that there are more differences between men and women in this case. Married rate is lower for lawyers in average and even lower for women, also the number of children is slightly lower for women. Although women head of the household are also less than men, this rate is higher for lawyers reaching 37% of them. 43 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 3 Summary Statistics: Law Male Obs Mean SD Labor Market Outcomes Hourly Wage Log(hourly wage) Monthly Wage Labor Market Participation Monthly Hours worked Real experience Real experience squared Mean of number jobs by year Level of responsibility Big Firm (>500w) Metropolitan Region Age Performance at University Reprove any class==1 Post graduate schooling==1 Family Background Mother's years of schooling Father's years of schooling Grades in secondary school Poor background==1 Uniparental home==1 Non Cognitive Abilities Self control test Self esteem test Family Status Number of children Married==1 Head of the household==1 Wage Intervals (1USD=537CHP) Less than 372 USD Between 372 and 745 USD Between 745 and 1120 USD Between 1120 and 1490 USD Between 1490 and 1862 Between 1862 and 2793 USD Between 2793 and 3725 USD Between 3725 and 4656 USD Between 4656 and 5587 USD More thab 5587 USD Total Obs Female Mean SD 182 11148.6 6598.35 183 7967.87 3765.2 182 9.17 0.57 183 8.84 0.63 182 2066832 1247710 183 1400567 645716 249 0.99 0.09 257 0.93 0.25 247 230.61 419.98 240 265.85 600.07 249 19.39 5.21 257 20.58 6.72 249 402.99 228.13 257 468.37 310.19 249 1.36 0.59 257 1.35 0.60 247 0.04 0.21 240 0.05 0.22 249 0.31 0.46 257 0.51 0.50 249 0.71 0.45 257 0.81 0.39 246 44.39 7.14 256 44.79 7.16 249 249 0.26 0.63 226 231 245 247 249 12.53 13.83 60.02 0.17 0.20 230 240 239 249 249 2 2 2 7 11 29 34 31 19 45 182 0.44 257 0.48 257 0.16 0.63 0.37 0.48 238 236 256 254 257 13.54 15.11 63.04 0.06 0.21 3.00 3.12 3.85 0.24 0.41 1.47 1.52 0.45 241 0.38 251 1.51 1.47 0.47 0.36 2.44 0.84 0.99 % 1.1 1.1 1.1 3.85 6.04 15.93 18.68 17.03 10.44 24.73 100 1.44 251 0.37 257 0.11 257 2.09 0.67 0.37 % 1.39 0.47 0.48 3.52 3.92 4.71 0.38 0.40 6 8 15 20 48 37 36 10 3 183 3.28 4.37 8.2 10.93 26.23 20.22 19.67 5.46 1.64 100 44 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 4 shows the summary of the descriptive statistics for the degree of Medicine by gender. In this case the gap in monthly wages is 76% approximately in favour of men. This is lower than in the case of business/economics and law. In addition, we can note that monthly hours worked by women are in average lower than hours worked by men and so the gap reduces to 91% in monthly hourly wage and to 99% if we look at the logarithm. We also can note that female labor force participation is 97% and is lower than male labor force participation which is 100%, both are higher than in case of business and law. The accumulated experience in terms of years of experience and number of parallel activities of women and men are the same. We can also observe that the proportion of doctors in job positions with more responsibility is nearly null for both gender. We can also observe that there are not great differences in the type of firm they work. In this case, 90% of women work in big firms and 86% of men. This latter statistics is higher than in the case of business and law. In addition, we can note that 97% of women and men that study medicine follow post graduate degrees. This latter may be related to obtaining of specialities. Again women have a slightly better performance at University and at school: a lower proportion of women reprove classes and they have higher grades at secondary school. Mother’s and father’s schooling are more similar among groups in this case and the level of parent’s schooling is higher in comparison to the other to professions. The measures of non cognitive abilities behave in the same way than the other to cases. In average women got a lower degree of self control but a higher degree of self esteem. It is worth noting that non cognitive abilities are higher in this professionals than in business and law. Finally, regarding the measures of current home situation present the following characteristics. We can see that medical professionals observed in this sample have less children than the other professional and women doctors have less children than men doctors. Married rate is lower for women than for men, however men have a higher married rate than the other two professions and women have higher married rate than lawyers but lower than business women. Although, again only 31% of women are head of the household in contrast to 99% of men. 45 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 4 Summary Statistics: Medicine Male Obs Mean SD Labor Market Outcomes Hourly Wage Log(hourly wage) Monthly Wage Labor Market Participation Monthly Hours worked Real experience Real experience squared Mean of number jobs by year Level of responsibility Big Firm (>500w) Metropolitan Region Age Performance at University Reprove any class==1 Post graduate schooling==1 Family Background Mother's years of schooling Father's years of schooling Grades in secondary school Poor background==1 Uniparental home==1 Non Cognitive Abilities Self control test Self esteem test Family Status Number of children Married==1 Head of the household==1 Wage Intervals (1USD=537CHP) Less than 372 USD Between 372 and 745 USD Between 745 and 1120 USD Between 1120 and 1490 USD Between 1490 and 1862 Between 1862 and 2793 USD Between 2793 and 3725 USD Between 3725 and 4656 USD Between 4656 and 5587 USD More thab 5587 USD Total Obs Female Mean SD 232 8046.97 5852.7 224 7303.7 4719.2 232 8.80 0.61 224 8.73 0.58 232 1171624 770749 224 889950 560867 255 1.00 0.06 256 0.97 0.16 254 152.33 54.64 249 144.61 250.10 255 13.24 2.63 256 13.34 2.47 255 182.07 94.40 256 184.05 65.73 255 1.24 0.46 256 1.22 0.44 254 0.01 0.11 249 0.00 0.06 255 0.86 0.34 256 0.90 0.30 255 0.59 0.49 256 0.77 0.42 254 38.55 3.33 256 38.75 2.73 255 255 0.16 0.97 242 241 254 253 255 13.35 14.42 64.41 0.21 0.13 240 244 1.29 1.33 248 255 255 2.02 1.24 255 0.87 0.33 256 0.99 0.09 256 % 1.29 3 8.62 19 7.33 35 8.19 32 12.93 46 30.17 55 16.81 18 5.6 8 3.02 3 6.03 5 100 224 3 20 17 19 30 70 39 13 7 14 232 0.37 256 0.17 256 0.20 0.97 0.40 0.17 252 252 254 256 256 13.93 15.04 65.98 0.12 0.13 3.26 3.84 2.03 0.32 0.34 0.44 241 0.36 253 1.32 1.29 0.41 0.32 3.72 3.94 2.93 0.41 0.33 1.92 1.34 0.74 0.44 0.31 0.46 % 1.34 8.48 15.63 14.29 20.54 24.55 8.04 3.57 1.34 2.23 100 46 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 4 The Results Descriptive statistics presented before help us to shed some light in which are the determinants of wages in the labor market. In this section we will use these measures to see whether once we take into account some of these differences we still have the gaps in wages we have in our data. As we pointed out in section 2 we are using two different specifications: OLS estimation and an ordered probit estimation. Tables 5, 6 and 7 present the results of the OLS regressions for each type of degree respectively. Regarding Business/Economics we can see that once the variables describe in the sections before are included the coefficient associated to the variable female start decreasing stadely until it turns to be not statistically significant in column 7. This latter column is the one including the vector of current family condition. This vector does not have a theoretical reason of why should be added in the wage equations however we added this variables in order to control for preferences of looking for certain types of jobs. Number of children and head of the household are positive and statistically significant. In fact, we know that being head of the household present additional responsibilities to finance household consumption. Other important variables which are determinants of business people’s wages are experience, the level of responsibility at the occupation, having a post graduate study and working in the metropolitan region. All these four variables add a premium on a professional’s wage in Business/Economics career. Regarding professionals in the Law degree we can note that just as before the coefficient associated to the dummy for female decreases steadily once different variables are added progressively until it turns not significant in column 7. In this case, only the number of children is a significant variable of the vector of current family conditions. However, this vector is picking up all the effect of gender. It is also new that in this wage equation the non cognitive ability test for self control is statistically significant. That is the higher the level of self control the higher are the wages. This is interesting and very intuitive to think, since lawyers need special abilities to be good professionals. As before real experience, the level of responsibility and a post graduate course helps to have higher wages. Regarding doctors we can note that female is never a negative issue in terms of wages. The only variables statistically significant in our regressions are the size of the firm in the sense 47 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E that the bigger the number of workers in the firm the lower is the wage, and working outside the metropolitan region give doctors higher wages. This latter may be due to scarcity of these professionals in the rest of the country as well as special government premiums to doctors working outside the metropolitan region. 48 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 5 OLS Regressions: Business/Economics, Dependent Variable=Log(Hourly Wage) (1) (2) (3) (4) (5) (6) Female==1 -0.259** -0.269** -0.260** -0.259** -0.231** -0.213** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Real experience 0.117** 0.111** 0.121** 0.090** 0.079** (0.000) (0.000) (0.000) (0.001) (0.007) Real experience squared -0.003** -0.003** -0.003** -0.002** -0.002* (0.000) (0.000) (0.000) (0.005) (0.043) Mean of number jobs by year 0.033 0.040 0.005 0.025 0.029 (0.808) (0.766) (0.970) (0.847) (0.830) Level of responsibility 0.109* 0.119* 0.123** 0.121** (0.021) (0.011) (0.007) (0.009) Big Firm (>500w) -0.023 -0.036 -0.009 -0.006 (0.614) (0.438) (0.845) (0.886) Metropolitan Region 0.170* 0.164 0.156 0.195* (0.048) (0.055) (0.056) (0.018) Post graduate schooling==1 0.110* 0.106* 0.107* (0.019) (0.020) (0.021) Reprove any class==1 -0.128 -0.105 -0.093 (0.051) (0.097) (0.144) Age -0.010 -0.009 -0.013 (0.200) (0.228) (0.084) Mother's years of schooling 0.003 0.006 (0.683) (0.485) Father's years of schooling 0.011 0.010 (0.170) (0.192) Grades in secondary school -0.004 -0.006 (0.579) (0.411) Poor background==1 -0.060 -0.077 (0.501) (0.399) Uniparental home==1 -0.016 -0.020 (0.798) (0.758) Self control test -0.064 (0.247) Self esteem test 0.010 (0.870) Married==1 (7) -0.090 (0.180) 0.060* (0.038) -0.001 (0.125) 0.041 (0.751) 0.108* (0.018) -0.024 (0.585) 0.196* (0.016) 0.114* (0.012) -0.102 (0.105) -0.014 (0.054) 0.006 (0.490) 0.009 (0.248) -0.003 (0.633) -0.118 (0.188) -0.015 (0.809) -0.036 (0.499) 0.011 (0.854) 0.085 (0.213) Number of children 0.051** (0.003) Head of the household==1 0.196** (0.004) Constant 9.327** 8.191** 8.054** 8.382** 8.641** 8.979** 8.703** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Observations 393 393 393 393 374 365 360 R-squared 0.072 0.128 0.150 0.174 0.181 0.190 0.244 p values in parentheses * significant at 5%; ** significant at 1% 49 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 6 OLS Regressions: Law, Dependent Variable=Log(Hourly Wage) (1) (2) (3) (4) (5) Female==1 -0.327** -0.304** -0.284** -0.326** -0.359** (0.000) (0.000) (0.000) (0.000) (0.000) Real experience 0.027 0.024 0.047 0.041 (0.322) (0.373) (0.096) (0.166) Real experience squared -0.001 -0.001 -0.001 -0.001 (0.120) (0.137) (0.153) (0.255) Mean of number jobs by year 0.054 0.056 0.057 0.028 (0.492) (0.482) (0.465) (0.742) Level of responsibility 0.105 0.097 0.063 (0.544) (0.571) (0.726) Big Firm (>500w) -0.084 -0.088 -0.088 (0.205) (0.181) (0.220) Metropolitan Region -0.050 -0.054 -0.110 (0.492) (0.456) (0.169) Post graduate schooling==1 0.109 0.102 (0.115) (0.168) Reprove any class==1 -0.045 -0.020 (0.586) (0.824) Age -0.021* -0.021* (0.010) (0.022) Mother's years of schooling 0.014 (0.304) Father's years of schooling 0.013 (0.308) Grades in secondary school -0.005 (0.604) Poor background==1 0.136 (0.267) Uniparental home==1 -0.146 (0.098) Self control test Self esteem test Married==1 Number of children Head of the household==1 Constant Observations R-squared p values in parentheses * significant at 5%; ** significant at 1% (6) -0.331** (0.000) 0.026 (0.376) -0.001 (0.337) 0.071 (0.423) 0.071 (0.691) -0.052 (0.472) -0.122 (0.129) 0.099 (0.186) -0.048 (0.600) -0.012 (0.196) 0.015 (0.270) 0.016 (0.205) -0.004 (0.719) 0.178 (0.148) -0.158 (0.079) -0.200** (0.007) 0.051 (0.599) (7) -0.234 (0.056) 0.001 (0.970) -0.000 (0.828) 0.090 (0.316) 0.072 (0.699) -0.050 (0.494) -0.116 (0.156) 0.126 (0.098) -0.048 (0.598) -0.013 (0.199) 0.013 (0.338) 0.015 (0.257) -0.003 (0.733) 0.144 (0.242) -0.147 (0.104) -0.194** (0.010) 0.062 (0.525) 0.058 (0.584) 0.069* (0.027) 0.103 (0.396) 9.168** 8.944** 9.035** 9.485** 9.602** 9.433** 9.393** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 365 365 365 362 315 299 297 0.069 0.085 0.091 0.122 0.165 0.182 0.202 50 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 7 OLS Regressions: Medicine, Dependent Variable=Log(Hourly Wage) (1) (2) (3) (4) (5) (6) Female==1 -0.067 -0.071 -0.019 -0.012 -0.024 -0.009 (0.225) (0.208) (0.733) (0.829) (0.684) (0.886) Real experience -0.032 -0.014 -0.026 -0.074 -0.081 (0.799) (0.910) (0.839) (0.562) (0.533) Real experience squared 0.002 0.001 0.001 0.003 0.003 (0.739) (0.848) (0.791) (0.580) (0.529) Mean of number jobs by year -0.062 -0.056 -0.055 0.019 0.002 (0.380) (0.419) (0.423) (0.790) (0.976) Level of responsibility 0.130 0.119 0.085 0.083 (0.662) (0.688) (0.773) (0.776) Big Firm (>500w) -0.355** -0.362** -0.344** -0.350** (0.000) (0.000) (0.000) (0.000) Metropolitan Region -0.182** -0.187** -0.176** -0.179** (0.002) (0.002) (0.004) (0.004) Post graduate schooling==1 -0.048 -0.115 -0.199 (0.781) (0.532) (0.305) Reprove any class==1 -0.059 -0.058 -0.074 (0.429) (0.438) (0.333) Age 0.002 0.011 0.007 (0.890) (0.495) (0.656) Mother's years of schooling 0.002 0.004 (0.853) (0.663) Father's years of schooling 0.011 0.009 (0.215) (0.339) Grades in secondary school 0.006 0.004 (0.607) (0.698) Poor background==1 -0.014 -0.032 (0.860) (0.684) Uniparental home==1 -0.089 -0.075 (0.292) (0.383) Self control test -0.081 (0.212) Self esteem test -0.040 (0.643) Married==1 (7) 0.109 (0.221) -0.087 (0.503) 0.003 (0.518) -0.009 (0.896) 0.096 (0.741) -0.378** (0.000) -0.162** (0.010) -0.218 (0.260) -0.047 (0.548) 0.005 (0.776) 0.003 (0.767) 0.010 (0.288) 0.005 (0.683) -0.045 (0.561) -0.079 (0.358) -0.073 (0.272) -0.036 (0.677) 0.062 (0.453) Number of children 0.045 (0.084) Head of the household==1 0.162 (0.075) Constant 8.801** 9.021** 9.304** 9.365** 8.813** 9.343** 9.219** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Observations 456 456 456 455 431 411 409 R-squared 0.003 0.005 0.072 0.075 0.075 0.081 0.097 p values in parentheses * significant at 5%; ** significant at 1% 51 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Tables 8, 9 and 10 present the results of the Ordered Probit regressions for each type of degree respectively. We think this model is more accurate because we did not have the real level of wages as a continuous variables for most of the sample. Regarding Business/Economics we can see that it is still the case that once the control variables are included the coefficient associated to the variable female decreases to turn into to zero in column 7. Again, the vector of current family conditions is driving this result. It is also maintained the conclusion that the other important variables are experience, the level of responsibility at the occupation, having a post graduate study and working in the metropolitan region. At the same time, performance at University and the self control non cognitive ability test are significant variables with the expected coefficients. Regarding professionals in the Law degree we can note that contrary to the case above women lawyers do have a cost in terms of wages because of being women. In this model, there is strong significance of metropolitan region and age which are negative, and the self control test is again statistically significant. Lawyers who have a better level of self control got higher wages. It is maintained also that the number of children and being had of the household are important variables in the wage equation whereas in this case this vector is not picking up the effect of gender. Regarding doctors the results are again very intuitive. We can observed that female turns to be a statistically not significant variable explaining wages. The size of the firm, working outside the metropolitan region and hours worked are statistically significant variables. It is also true that self control and family conditions are also statistically significant variables. 52 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 8 Ordered Probit Regressions: Business/Economics, Dependent Variable=Wage Intervals (1) (2) (3) (4) (5) (6) Female==1 -0.867** -0.924** -0.926** -0.938** -0.698** -0.643** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Monthly Hours Worked 0.000 0.000 0.000 0.000 0.011** 0.012** (0.417) (0.161) (0.389) (0.292) (0.000) (0.000) Real experience 0.286** 0.273** 0.303** 0.263** 0.238** (0.000) (0.000) (0.000) (0.000) (0.001) Real experience squared -0.006** -0.006** -0.006** -0.006** -0.005* (0.000) (0.000) (0.000) (0.001) (0.013) Mean of number jobs by year -0.257 -0.261 -0.356 -0.145 -0.061 (0.408) (0.402) (0.256) (0.665) (0.857) Level of responsibility 0.476** 0.500** 0.414** 0.416** (0.000) (0.000) (0.000) (0.000) Big Firm (>500w) 0.262* 0.234* 0.155 0.164 (0.016) (0.033) (0.176) (0.157) Metropolitan Region 0.345 0.344 0.324 0.453* (0.086) (0.088) (0.116) (0.031) Post graduate schooling==1 0.263* 0.297** 0.283* (0.017) (0.010) (0.016) Reprove any class==1 -0.354* -0.356* -0.319 (0.023) (0.027) (0.050) Age -0.015 -0.017 -0.030 (0.423) (0.367) (0.118) Mother's years of schooling 0.000 0.008 (0.997) (0.710) Father's years of schooling 0.036 0.036 (0.073) (0.069) Grades in secondary school -0.012 -0.019 (0.487) (0.279) Poor background==1 -0.255 -0.262 (0.248) (0.244) Uniparental home==1 -0.054 -0.068 (0.732) (0.668) Self control test -0.326* (0.018) Self esteem test 0.019 (0.902) Married==1 Number of children Head of the household==1 Observations p values in parentheses * significant at 5%; ** significant at 1% 393 393 393 393 374 365 (7) -0.341 (0.053) 0.012** (0.000) 0.210** (0.005) -0.004* (0.029) -0.033 (0.923) 0.402** (0.001) 0.137 (0.241) 0.467* (0.027) 0.312** (0.009) -0.356* (0.032) -0.033 (0.083) 0.007 (0.735) 0.035 (0.083) -0.015 (0.394) -0.387 (0.092) -0.063 (0.693) -0.275* (0.048) 0.028 (0.858) 0.137 (0.435) 0.108* (0.018) 0.526** (0.002) 360 53 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 9 Ordered Probit Regressions: Law, Dependent Variable=Wage Intervals (1) (2) (3) (4) (5) (6) Female==1 -0.723** -0.710** -0.681** -0.792** -0.935** -0.899** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Monthly Hours Worked 0.000 0.000 0.000 0.000 0.000 0.000 (0.136) (0.136) (0.136) (0.155) (0.155) (0.067) Real experience 0.052 0.047 0.102* 0.107* 0.083 (0.262) (0.316) (0.036) (0.035) (0.114) Real experience squared -0.002 -0.001 -0.001 -0.001 -0.001 (0.140) (0.166) (0.210) (0.276) (0.369) Mean of number jobs by year -0.071 -0.065 -0.054 -0.136 -0.051 (0.599) (0.630) (0.690) (0.361) (0.745) Level of responsibility 0.418 0.409 0.285 0.302 (0.163) (0.176) (0.368) (0.341) Big Firm (>500w) -0.030 -0.033 -0.029 0.034 (0.790) (0.775) (0.816) (0.790) Metropolitan Region -0.167 -0.189 -0.299* -0.323* (0.181) (0.136) (0.031) (0.023) Post graduate schooling==1 0.151 0.120 0.118 (0.208) (0.349) (0.369) Reprove any class==1 -0.088 -0.033 -0.100 (0.536) (0.829) (0.530) Age -0.057** -0.059** -0.044** (0.000) (0.000) (0.009) Mother's years of schooling 0.040 0.040 (0.081) (0.086) Father's years of schooling 0.022 0.028 (0.317) (0.201) Grades in secondary school 0.007 0.010 (0.690) (0.559) Poor background==1 0.183 0.257 (0.391) (0.238) Uniparental home==1 -0.235 -0.273 (0.128) (0.087) Self control test -0.326* (0.012) Self esteem test 0.160 (0.347) Married==1 Number of children Head of the household==1 Observations p values in parentheses * significant at 5%; ** significant at 1% 365 365 365 362 315 299 (7) -0.568** (0.008) 0.000 (0.088) 0.015 (0.785) 0.000 (0.776) -0.010 (0.948) 0.435 (0.193) 0.026 (0.841) -0.355* (0.014) 0.166 (0.216) -0.109 (0.500) -0.047** (0.006) 0.033 (0.166) 0.030 (0.186) 0.009 (0.587) 0.163 (0.456) -0.283 (0.079) -0.351** (0.008) 0.184 (0.285) -0.002 (0.991) 0.170** (0.002) 0.463* (0.028) 297 54 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 10 Ordered Probit Regressions: Medicine, Dependent Variable=Wage Intervals (1) (2) (3) (4) (5) (6) Female==1 -0.401** -0.412** -0.315** -0.312** -0.120 -0.095 (0.000) (0.000) (0.002) (0.002) (0.283) (0.407) Monthly Hours Worked 0.001** 0.001** 0.001** 0.001** 0.013** 0.013** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Real experience -0.064 -0.051 -0.056 -0.066 -0.070 (0.769) (0.818) (0.801) (0.776) (0.773) Real experience squared 0.004 0.003 0.003 0.003 0.003 (0.646) (0.695) (0.694) (0.766) (0.733) Mean of number jobs by year -0.265* -0.259* -0.270* -0.021 -0.060 (0.030) (0.036) (0.029) (0.875) (0.654) Level of responsibility 0.541 0.559 0.074 0.053 (0.301) (0.287) (0.890) (0.922) Big Firm (>500w) -0.522** -0.521** -0.591** -0.637** (0.001) (0.001) (0.000) (0.000) Metropolitan Region -0.426** -0.429** -0.380** -0.385** (0.000) (0.000) (0.001) (0.001) Post graduate schooling==1 0.112 -0.125 -0.371 (0.714) (0.711) (0.304) Reprove any class==1 0.065 -0.057 -0.100 (0.627) (0.678) (0.482) Age 0.002 0.025 0.019 (0.939) (0.407) (0.548) Mother's years of schooling 0.007 0.012 (0.696) (0.525) Father's years of schooling 0.018 0.015 (0.282) (0.395) Grades in secondary school 0.002 -0.002 (0.910) (0.924) Poor background==1 -0.011 -0.053 (0.937) (0.715) Uniparental home==1 -0.156 -0.141 (0.317) (0.380) Self control test -0.306* (0.011) Self esteem test -0.057 (0.726) Married==1 Number of children Head of the household==1 Observations p values in parentheses * significant at 5%; ** significant at 1% 456 456 456 455 431 411 (7) 0.214 (0.201) 0.013** (0.000) -0.076 (0.754) 0.003 (0.740) -0.087 (0.520) 0.094 (0.862) -0.722** (0.000) -0.349** (0.003) -0.407 (0.261) -0.045 (0.758) 0.014 (0.668) 0.009 (0.658) 0.018 (0.308) -0.002 (0.926) -0.082 (0.573) -0.152 (0.347) -0.301* (0.016) -0.053 (0.744) 0.076 (0.624) 0.101* (0.038) 0.445** (0.009) 409 55 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 5 Conclusions This paper study differences in wages of three types of professionals in Chile: Business women and men, lawyers and doctors. Our preferred specification is an ordered probit model. In this specification we can see that female does seem to have only a negative effect on wages for lawyers, even including current family conditions. Business women and men differences disappear once the vector of current family conditions are added. And doctors seems to have no differences in wages due to gender. Other important variables explaining differences in wages are the level of responsibility in the job, having postgraduate studies, the size of the firm, a regional effect. And most importantly, there is an important and positive effect of the non cognitive ability test that measures self control. 56 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E References Altonji, Joseph and Rebecca Blank (1999). “Race and Gender in the Labor Market.” Handbook of Labor Economics, 3, pp. 3143-3259. Becker, Gary (1971) The Economics of Discrimination, 2nd Edition, The University of Chicago Press, IL. Becker, Gary (1991) A Treatise on the Family, enlarged edition: Harvard University Press, Cambridge, MA. Blinder, Alan (1973). “Wage Discrimination: Reduced Form and Structural Estimates.” The Journal of Human Resources, 7(4), pp. 436-55. Contreras, Dante y Esteban Puentes (2001) “Is The Gender Wage Discrimination Decreasing In Chile? Thirty Years Of ‘Robust’ Evidence”, Departamento de Economía, Universidad de Chile. Fernandez, Fogli And Olivetti (2004) “Preference Formation And The Rise Of Women’s Labor Force Participation: Evidence From WWII”, NBER Working Paper 10589 Heckman, James (1998). “Detecting Discrimination.” The Journal of Economic Perspectives,12(2), pp. 101-116. Heckman, James, Jora Stixrud and Sergio Urzua (2005) “The Effects of Cognitive and Noncognitive Abilities on Labor Market Outcomes and Social Behavior”, University of Chicago. Montenegro, Claudio (1999) “Wage distribution in Chile: Does Gender Matter? A Quantile Regression Approach. Mimeo, Universidad de Chile. Montenegro, Claudio y Paredes, Ricardo (1999) “Gender Wage Gap and Discrimination: A Long Term View Using Quantile Regression”. Mimeo, Universidad de Chile. Neal, Derek A. and William R. and Johnson (1996) “The Role of Premarket Factors in Black-White Wage Differences”, The Journal of Political Economy, Vol. 104, No. 5 (Oct., 1996), 869-895. Nuñez, Javier and Roberto Gutierrez (2004) “Classism, Discrimination and Meritocracy in the Labor Market: The Case of Chile.” Documento de trabajo 208. Departamento de Economia, Universidad de Chile. 57 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Oaxaca, Ronald (1973). “Male-Female Wage Differentials in Urban Labor Market.” International Economic Review, 14(3), pp. 693-709. O’Neil, June E. and Dave M. O’Neil (2005) “What do wage Differentials tell us about labor Market Discrimination”, NBER Working Paper 11240. Paredes, Ricardo y Riveros, Luis (1994): “Gender Wage Gaps in Chile. A Long term View:1958:1990”. Estudios de Economía, Vol.21, Número especial, 1994. Rosenberg, M. (1965). Society and the Adolescent Self-Image. Princeton, NJ: Princeton University Press. Rotter, J. B. (1966). Generalized Expectancies for Internal versus External Control of Reinforcement. Washington DC: American Psychological Association. 58 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Appendix to Chapter 2 This survey was applied to ex-alumni of the Universidad de Chile that studied Medicine, Law and Business/Economics and who graduated at least eight years ago. The sample is made up of 50% men and 50% women. The survey was implemented by telephone. To offer the respondents an image of the survey, a website has been designed to present the Survey, which provides a description of the survey, its objectives and questionnaires. The survey takes approximately 20 minutes and contains six modules: General Information, Education, Employment History, Family Background, Individual History and Test of NonCognitive Abilities. A1. Calendar of Activities The following table shows the Calendar of Activities developed for the implementation of the Survey: Date December, 2005January, 2006 20th January 10th February March, 2006 April, 2006 May, 2006 30th May June, 2006 Activity Design of the Questionnaire Sample framework: to locate the address of the students in the university records Progress Report and Work Plan Videoconference Design of the Questionnaire Sample framework: to locate the address of the students in the university records Design of the Questionnaire Sample framework: to locate the address of the students in the university records Pilot Survey Final questionnaire Questionnaire Manual for Interviewers Interviewer Training Survey Starts Data Entry and Validation of the Survey Starts First draft Survey Continues Data Entry and Validation of the Survey Continues 59 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 20th June July, 2006 6th September 20th October 29th November A2. Workshop Survey Continues Data Entry and Validation of the Survey Continues Second draft Data Entry and Validation of the Survey Continues Final Workshop Final version Sample Design The survey is being developed without geographic restrictions. Since it is a telephone based survey, there are no geographic boundaries. It is simply a case of locating the individuals of the sample in the city where they may be. The selection process of the sample was developed in the following stages. A3. Search for Names of Ex-alumni First, a search was made for administrative information on ex-alumni in the Faculties of the Universidad de Chile and centrally. We had several meetings with Central authorities of the University who finally accepted to furnish us with a database of graduates of the University from 1970 to 1997 from the three degree programs involved. This database is confidential and contains the national identification number of the person, their name, year of graduation, degree program and, in some cases, an address. Table 1 shows the distribution of the original framework sample, obtained from the administrative information of the Universidad de Chile. 60 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Year 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Total A4. Female 27 14 20 28 61 37 41 20 28 35 26 52 72 69 21 41 36 34 59 46 89 80 85 49 51 46 32 46 1245 Economics Male 90 74 98 98 186 123 168 112 83 152 99 132 189 191 98 182 125 103 98 97 140 136 109 53 94 67 74 84 3255 Ta ble 1 : Origin a l Fr a m e w or k Sa m ple Law Female Male Sub total Sub total 117 56 153 209 88 41 139 180 118 50 132 182 126 56 119 175 247 44 113 157 160 36 135 171 209 52 107 159 132 30 68 98 111 27 91 118 187 27 106 133 125 58 165 223 184 64 111 175 261 42 118 160 260 53 112 165 119 72 120 192 223 43 137 180 161 46 107 153 137 28 85 113 157 30 87 117 143 32 84 116 229 28 93 121 216 23 78 101 194 39 115 154 102 52 133 185 145 45 115 160 113 62 106 168 106 87 125 212 130 76 140 216 4500 1299 3194 4493 Female 36 42 48 41 50 80 90 115 138 206 98 165 125 142 123 115 74 80 89 103 111 98 86 88 97 72 91 87 2690 Medicine Male 181 164 158 158 178 209 165 240 231 376 156 233 204 273 226 235 152 212 159 167 161 155 131 147 168 133 132 132 5236 Total Sub total 217 206 206 199 228 289 255 355 369 582 254 398 329 415 349 350 226 292 248 270 272 253 217 235 265 205 223 219 7926 543 474 506 500 632 620 623 585 598 902 602 757 750 840 660 753 540 542 522 529 622 570 565 522 570 486 541 565 16919 Updating Ex-alumni Data The addresses and other personal data on ex-alumni obtained from the administrative data showed a significant proportion of incomplete records with outdated information. In order to update the original information, 6,000 individuals were chosen, who were tracked down in phone books and other sources to get their location data. After this search process, the following distribution was obtained (See Table 2). This will finally be the real sample framework from which the final sample is chosen. Year 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Total Female 69 21 41 36 34 59 46 89 80 85 49 51 46 32 46 784 Economics Male 191 98 182 125 103 98 97 140 136 109 53 94 67 74 84 1651 Ta ble 2 : Re a l Sam ple Fra m e w or k Law Female Male Sub total Sub total 42 118 160 260 53 112 165 119 72 120 192 223 43 137 180 161 46 107 153 137 28 85 113 157 30 87 117 143 32 84 116 229 28 93 121 216 23 78 101 194 39 115 154 102 52 133 185 145 45 115 160 113 62 106 168 106 87 125 212 130 76 140 216 2435 758 1755 2513 Female 111 98 86 88 97 72 91 87 730 Medicine Male 161 155 131 147 168 133 132 132 1159 Total Sub total 272 253 217 235 265 205 223 219 1889 160 425 311 403 314 250 274 259 622 570 565 522 570 486 541 565 6837 61 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E A5. Selection of the Sample The definitive sample is chosen based on the real sample framework defined in the point above. The objective number of surveys for carrying out is 1,500. One third of these correspond to each degree program, and in equivalent proportions between men and women. In order to effectively obtain the surveys requested, it is necessary to have an over-sizedsample, to be able to cover the losses arising from people that cannot be found or that refuse to participate in the survey. Based on earlier studies and considering the lack of individual information available, we can consider a loss of 100%. Therefore, the selected sample should be 3,000 individuals. The selected sample is obtained by choosing 1,000 individuals graduated in each of the three degree programs (Law, Medicine and Economics) randomly. The same number of men and women are chosen within each degree program. To complete the sample, by degree program, the same number of male and female graduates by graduation year are chosen. Therefore, the final sample, displayed in Table 3, may be characterized as probabilistic, stratified by degree programs and gender, with a nonproportional distribution among strata. Year 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Total A6. Female 46 89 80 85 49 51 46 32 46 524 Economics Male 60 86 83 67 33 57 42 45 51 524 Sub total 106 175 163 152 82 108 88 77 97 1048 Table 3: Final Sample Law Female Male Sub total 28 37 65 30 38 68 32 36 68 28 40 68 23 34 57 39 50 89 52 57 109 45 50 95 62 46 108 87 54 141 76 60 136 502 502 1004 Female 86 88 97 72 91 87 521 Medicine Male 81 91 104 82 81 82 521 Total Sub total 167 179 201 154 172 169 1042 65 68 174 243 220 408 370 404 350 390 402 3094 Pilot survey Before implementing the survey, a pilot survey was carried out on the whole sample selected. The general objective of this pilot survey is to evaluate the operation of the questionnaire by means of a telephone interview. It also has the following specific objectives: 62 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 1. Review problems of content (difficulty of comprehension on the part of the respondents, lack of response categories, etc.). 2. Evaluate the implementation periods. 3. Difficulty in contacting and locating respondents. To carry out the Pre-Test, a sample of graduates that were not included in the selected sample were extracted, from 70 cases of each of the degree programs chosen for the study. These 70 cases were in turn divided evenly among men and women. Table 4: Sample Pre-Test Degree program Law Medicine Economics Total Men Women 35 35 35 105 35 35 36 106 Total 70 70 71 211 The Field Coordinator and the Survey Programmer were responsible for the training of the telephone operators that carried out the pilot survey. The training consisted of a presentation of the study, which was followed by a review of the questionnaire. It was carried out in the morning of the first day of work of the operators After the end of the pilot survey, the questionnaire was modified slightly to gather the observations made through the implementation. A7. Questionnaire and Interviewer Manual The Survey is designed for telephone as well as paper based implementation, in case an interviewer should have to implement it so. The Questionnaire that will finally be implemented is presented in the Appendix to this chapter of the Report and is comprised of 5 modules of questions and two non-cognitive ability tests that are to be found at the end there. The form covers areas such as: household structure and identification, income, job, education, health, housing, family background and perceptions. The modules are as follows: • Module A: General Information of the Respondent Objective: Obtain information on sex, marital status, age and position within the household. 63 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E • • • • • • Module B: Education Objective: Obtain Information on prior education of the respondent and also on activities subsequent to university. Questions are posed on the quality of the secondary education received. Module C: Employment History Objective: Obtain complete information on the respondents employment activities from their date of graduation. They are also questioned on their parallel activities and job characteristics. For those who are currently inactive, questions are posed to obtain information on the reserve salary. This allows us to discover the real employment experience of men and women. Module D: Family Background Objective: Obtain information about the parents’ education and the emotional and socioeconomic stability of the household during childhood. There are also questions on the size of household, gender composition and education level of siblings. Module E: Personal History Objective: Obtain information on respondent’s marital history and common-law wives, as well as children. Test 1: Rotter Internal-External Locus of Control Scale It is a four-item abbreviated version of a 23-item forced choice questionnaire adapted from the 60-item Rotter scale developed by Rotter (1966). The scale is designed to measure the extent to which individuals believe they have control over their lives, i.e., self-motivation and self-determination, (internal control) as opposed to the extent that the environment (i.e., chance, fate, luck) controls their lives (external control). The scale is scored in the internal direction: the higher the score, the more internal the individual. Individuals are first shown two sets of statements and asked which of the two statements is closer to their own opinion. They are then asked whether that statement is much closer or slightly closer to their opinion. These responses are used to generate four-point scales for each of the paired items, which are then averaged to create one Rotter Scale score for each individual. Test 2 Rosenberg Self-Esteem Scale It is a 10-item scale, designed for adolescents and adults; measures an individual’s degree of approval or disapproval toward himself (Rosenberg, 1965). The scale is short, widely used, and has accumulated evidence of validity and reliability. It contains 10 statements of self-approval and disapproval to which respondents are asked to strongly agree, agree, disagree, or strongly disagree. 64 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E A8. Preparation of the Survey Fieldwork preparation requires carrying out all the regular tasks, in other words, registration, training, supervision, as well as preparing and providing the necessary material and inputs for survey implementation. The selection method for the interviewers was by invitation. These invitations were made to interviewers that have worked in other similar surveys undertaken previously by the Centro de Microdatos. In fact, 10 telephone interviewers were invited, who possessed previous training in the same characteristics as the Pilot survey. In this particular occassion, they also received an Interviewer Manual. All operators who implemented the survey have higher education studies, either technical level or university. The training activity took approximately 4 hours. All the questions of the questionnaire were reviewed, and the concepts required to implement it were defined, as well as the aspects that had to be emphasized in the survey, in addition to the clarification of any pertinent queries. A product of this stage was the Interviewer Manual with all the final corrections. A9. Organization of the Work Team The fieldwork team is finally composed of: • A technical coordinator of the Survey, responsible for ensuring the correct implementation of the methodology and quality standards. He/she is responsible for verifying the correct implementation in the field, fulfillment of the sample sizes, and the subsequent verification of the control tabulations for the final approval of the database. • • • A logistics and control coordinator, responsible for the correct execution and control of the administrative and financial processes. Responsible for monitoring the state of progress and ensuring observance of the work calendar. A field work coordinator, responsible for distributing the sample among the telephone operators and supervising the work carried out by them. An I.T. coordinator, responsible for designing, implementing and administering the information systems for monitoring field work, data entry of surveys, data validation and structuring the final magnetic file. 65 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E • A sample designer, responsible for creating the sample design and subsequent calculation of the expansion factors. Ernesto Castillo. The Centro de Microdatos was responsible for preparing all the necessary inputs for the implementation of the survey, training classrooms, telephones, offices, office supplies, manuals and forms, transport and personnel. Finally, the Survey is final database for analysis was available on October 25th. 66 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Chapter 3 “Ability, Schooling Choices and Gender Labor Market Discrimination: Evidence for Chile” Abstract This paper presents a comprehensive analysis of the gender differences in the Chilean labor market. We formally deal with the selection of the individuals into schooling levels and its consequences on the gender gaps. Our approach allows for the presence of not only heterogeneity in observable variables but also unobserved heterogeneity. We link this unobserved heterogeneity to unobserved scholastic ability. In the analysis, we utilize a new and rich data set for Chile. This data set contains information on labor market outcomes (including labor history), on schooling attainment and schooling performance, and on a complete set of variables characterizing the family background of the individuals in the sample Our results show that there exist statistically significant gender differences in several dimensions of the Chilean labor market. Nevertheless, we show that these gaps critically depend on the schooling level of the individuals considered in the analysis. For example, the results indicate that there are no gender differences in labor market variables among college graduates (except in the case of hourly wages). We interpret our results with prudence. Specifically, instead of interpreting our findings as decisive evidence of the existence of discrimination in the Chilean labor market, we argue that future research based on better information might indeed explain some of the unexplained labor market gaps presented in this paper. In this context, our results represent a new and important attempt to provide a full understanding of the structural causes of gender gaps in the Chilean labor market but they are not conclusive. 67 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 1 Introduction Gender gaps in a variety of labor and educational outcomes (e.g. wages, earnings, employment, schooling levels) are well documented. The structural reasons behind these gaps, however, are not fully understood. This paper contributes to the literature by studying gender differences in a framework in which schooling decisions and labor market outcomes are endogenously determined. Our framework also allows individual heterogeneity not only from the point of view of observable characteristics but also unobserved variables. We assume that individuals know this additional source of heterogeneity, and they base their schooling and labor market decisions on it. Unobserved heterogeneity plays a crucial role in our approach. Ours is a challenging task for several reasons. First, a comprehensive analysis of gender differences in a variety of outcomes is subject to the usual and irremediable data limitations. Second, the natural complexity associated with econometric models of multiple, endogenous, and correlated outcomes makes these models usually not very empirically appealing. And finally, the fact that we allow individuals decisions to depend on variables unobserved by the researcher but known to the agent represents an additional challenge of our approach. Nevertheless, we deal with each of these difficulties. First, we utilize a new data set from Chile that contains detailed information on labor market and schooling outcomes at the individual level. Second, we postulate a simple factor structure model based on economic theory that simplifies the manner we can deal with multiple endogenous variables. And finally, we interpret this factor as unobserved heterogeneity since the researcher does not need to know the individual factor (although it is assumed to be known by the individual). We argue that the factor represents a combination of different scholastic skills (cognitive and noncognitive skills). As previously mentioned, we implement our approach using new information from Chile. The Chilean case provides an interesting example of apparently huge gender gaps in different dimensions of the labor market. Table 1 presents basic information for a variety of schooling and labor market outcomes obtained from a sample of males and females with ages between 28 and 40 years. 25 The evidence in Table 1 provides an initial flavor of the gender differences that motivate the idea of this paper. A comparison of the schooling outcomes (Panel A 25 The information comes from the Social Protection Survey 2002 of Chile (SPS02) which is the source of information used in this paper. This survey is described in detail in Section 2. 68 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E in Table 1) leads to conclude that, in average, (i) women are slightly more educated than males, (ii) women are less likely to repeat a grade in both primary and secondary school, and (iii) women show a better performance in school than males (measured by the average grade in secondary school). However, this educational advantage of women over men seems to have no consequences on the labor market. The evidence in Panel B illustrates this point. It shows that males overwhelmingly dominate females in every single dimension of the labor market (monthly earnings, employment, and experience). This paper studies the factors explaining these gender differences in labor market and schooling outcomes. The paper is organized as follows. Section 2 describes the data. Section 3 presents evidence on the differences in labor market outcomes between males and females using a conventional approach. Section 4 introduces our model and discusses its empirical implementation. Section 5 presents a discussion of our results. Section 6 concludes. 2 Data This paper uses information from the Chilean Social Protection Survey 2002 (SPS02). This survey was designed to identify and analyze the most important determinants of the social security decisions (participation in the social security system) among Chileans. In order to do this, a representative sample of 17,246 participants of the Chilean pension system was interviewed between June of 2002 and January of 2003. For each individual in the sample, the survey collected information on household composition (ages, genders and schooling levels of the household members as well as their relations with the interviewee), current employment status, different sources of income, schooling (maximum schooling attained, average grades in primary and secondary school, characteristics of the primary and secondary school attended), family history (mother's and father's education, characteristics of the place of residence where the individual grew up, and number of previous relationships), labor history since age 15 or since 1980 depending on the year the individual became 15 years old (periods of employment, unemployment and inactivity), training programs (information on the three most important training programs since 1980), expectations (job, retirement and life), savings (instruments and amounts), and a set of variables describing the individual's knowledge of the characteristics and performance of the Chilean pension system. 69 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E We use a sample of individuals with ages in the range of 28 and 40 years. This group represents approximately the 21% of the original sample (3,566 versus 17,246). 26 We restrict the ages of sample for several reasons. First, since the information on labor history begins only in 1980 (or since age 15), by using individuals 28-40 years old we assure that our sample report complete labor histories from age 18. Second, since schooling is an important ingredient of our analysis, by excluding individuals 27 years old and younger we focus our attention on individuals that have most likely reached their final schooling level. 27 Finally, it is worth noting that the current Chilean schooling system was designed only in the early 80s. Therefore, since our analysis includes information on the characteristics of the primary and secondary schools in which the individual was enrolled, by restricting the analysis to the individuals with ages 28-40, we assure that such information is available for most of our sample. Table A.1 presents the summary statistics of the variables used in this paper. 3 The Conventional Gender Gap Analysis The gender differences in labor market outcomes are usually analyzed in the context of linear models in which the variable of interested is regressed on the gender dummy variable and set of additional controls. The coefficient associated with the gender dummy is interpreted as the estimated gender gap. Given its popularity, our first attempt to quantify gender gaps follows closely this idea. Table 2 presents the results from the following model of (log) hourly wages (lnW): 26 Our sample is obtained after considering the following exclusions. We first exclude the military sample (57 individuals) and individuals reporting as occupation "family member without salary" (12 individuals). Then, we exclude individuals 27 years old or younger and 41 years old or older. With this the sample reduces from 17,177 to 5,439. Finally, individuals with missing values in any of the following variables are excluded: "years of education", "mother's education", "father's education", "growing up in poverty" and "growing up in a single parent household". This exclusion reduces the sample to the final 3,566 individuals. It is worth noting that the final exclusion is required since for each individual we need to have valid values for the controls entering in the schooling decision model presented in Section 3.1. 27 A more general analysis of the schooling decisions would require a dynamic model for schooling choices. The SPS02 does not allow us to carry out such analysis. 70 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E ln W = α + ϕ Gender + β X + U (1) where Gender represents the gender dummy (Gender=1 if individual is Male and 0 if Females), X represents individual's observable characteristics, and U is the error term in the regression. In this simple model, the (conditional) gender gap is simply ϕ. Each column in Table 2 represents a different specification of (1). In particular, column (A) presents the results of a model in which we include the characteristics of both the place of residence and occupation in the vector of controls X. Column (B) adds a set of variables controlling for the individual's accumulated experience and column (C) adds to the controls in (B) a set of variables controlling for schooling levels. The results indicate that males make approximately 23% more than females in terms of hourly wages. This gender gap is statistically significant regardless of the column analyzed. The last model in Table 1 (column D) includes a correction for the fact that the labor market outcome is reported only for individuals working (Heckman, 1974). This is particularly important given the gender differences in employment rates reported in Table 1 (panel B). Thus, the model in column D is: ln W = α + ϕ Gender + βX + U if wage is observed (D=1) D = 1[γZ + V > 0] (2) where 1[A] is an indicator function that takes a value of 1 if A is true and zero otherwise, Z is a vector of observables and V represents the unobservables. D=1[.] is the censoring rule for wages. In Z we include variables such as number of children, whether or not the individual grew up in a poor household, mother’s and father’s occupational status. The estimated gap after correcting for selection is 29% and it is statistically significant. Thus, after controlling for selections, we not only find a significant but larger gender gap in wages (compare to the ones estimated without using the correction). This fact illustrates the importance of paying particular attention to individual's endogenous decisions (in this case employment decisions) when analyzing gender gaps. We exploit this point in the following section. The analysis of gender gaps in wages is interesting and important but it represents only one dimension of many among which males and females can differ. We first extend our analysis to the case of monthly hours worked. We model (log) hours worked using a linear-in-parameter models similar to (1) and the same set of controls as the ones utilized for wages. Table 3 presents the estimates of gender gaps in this case. The structure of this table is identical to the one in Table 2. The 71 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E results from columns (A), (B) and (C) suggest that males work approximately 11% more hours per month than females. This difference is statistically significant and it is stable across the three specifications. However, the last column in Table 3 presents (again) a different story. Unlike the results for wages, the correction for selection significantly reduces the gender gap in hours worked. The estimated gap is only 0.04% and it is not statistically significant. We also extend our analysis to employment status. In this case, we use a probit model instead of a linear regression model. Table 4 presents the results for three different specifications. For each specification, we present both estimated coefficients and estimated marginal effects. 28 The results indicate that males are 22% more likely to report an employment (during the month previous to the date of the interview) than females when schooling and experience are excluded as controls. When schooling or schooling and experience are included as controls the estimated gap is 14%. The gaps are statistically significant regardless of the specification. In summary, the results show that men dominate women in every labor market outcome. Additionally, the results are robust across different specifications and only in the case of hours worked and after controlling for selection we find neither sizeable nor statistically significant gender differences. Notice that up this point we have utilized the individual's schooling decisions and accumulated experience as exogenous regressors. However, in principle these variables can also be subject to gender differences. Tables 5 and 6 present evidence on this point. The implications of separate analyses of schooling choices and accumulated experience on our previous results are left for the next section where they are discussed in the context of a more general framework than the one used here. 29 We model accumulated experience assuming that, whatever experience level is observed in the sample, it is the result of a decision involving three alternatives: less than 10 years of experience, between 10 and 15 years of experience, and more than 15 years of experience. This decision is assumed to depend on the schooling level of the individual as well as on his family background (mother's and father's education, broken home, age, and growing up in poverty). Given this set up, we compute the gender gaps in accumulated experience by estimating a multinomial probit model. Table 5 presents the estimated coefficients and marginal effects. 28 The marginal effects are computed at the mean values of the variables in the model. This is particularly important if we consider that schooling decisions and accumulated experience are probably endogenous variables in the context of the models presented in Tables 2, 3 and 4. The model presents in the next section deals with this possibility. 29 72 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E The estimates associated with the gender dummy are all significant and suggest that males are considerably more likely to report more experience than females. Specifically, males are 40% less likely to report less than 10 years of experience and 29% more likely to report more than 15 years of experience than females. The analysis of gender differences in schooling decisions is also relevant in the context of the previous results. On the one hand, if males are in fact more likely to report higher schooling levels than females (after controlling for observable characteristics), then the gender differences in labor market outcomes (including accumulated experience) could be simply interpreted as the result of gender differences in accumulated human capital. On the contrary, if females are more likely to report higher schooling levels than males, then the estimated gender differences in labor market outcomes could be interpreted as downward biased estimates of the actual gaps. Table 6 sheds light on existence of gender gaps in schooling decisions. It presents the coefficients and marginal effects obtained from a multinomial schooling choice model. The model is estimated using the maximum schooling levels reported by the individuals in the sample. The schooling levels considered are: primary school, secondary school, some post-secondary education, and complete tertiary education (college graduates). The results show that (if anything) females are more likely than males to reach higher schooling levels. The advantage of females over males in schooling achievement/attainment is confirmed in Table 7. This table presents the estimated gender gaps for three variables measuring schooling performance: probability of a grade repeated during primary school, probability of a grade repeated during secondary school, and average grades during secondary school. For each variable we consistently observe that females outperform males. Males are 7% and 4% more likely to repeat a grade during primary and secondary school, respectively, and males in average have a significantly lower grades during high school than females (0.31 points of test’s standard deviation). Therefore, the evidence presented in Tables 6 and 7 leads us to conclude that females should be better prepared than males to face the labor market. This also implies that by not including the gender differences in schooling variables our previous results might be underestimating the actual unexplained gender gaps (or discrimination). We analyze this possibility by introducing a more general model in which schooling decisions, schooling achievement, employment decisions, accumulated experience, hours worked and hourly wages are modeled jointly. 73 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 4 A Model of Schooling and Labor Market Outcomes under Unobserved Heterogeneity The model in this section follows the analysis in Heckman, Stixrud and Urzua (2006). Heckman, Stixrud and Urzua (2006) postulate and estimate a model with two underlying sources of unobserved heterogeneity that they interpreted as abilities (cognitive and noncognitive abilities). Conditioning on the observables, these factors account for all of the dependence across choices in the model. They show that both abilities play a crucial role explaining a variety of labor market and behavioral outcomes. In this paper we postulate the existence of only one underlying factor representing unobserved heterogeneity. This is mainly due to the fact that we do not have a set of cognitive and noncognitive variables in the SPS02 sample. Consequently, we interpret the source of unobserved heterogeneity as a combination of both cognitive and noncognitive abilities. 30 The identification of its distribution is discussed in Section 3.4 below. Let θ denote the unobserved heterogeneity or latent ability. We assume this latent ability determines the individual's schooling and labor market outcomes, and that there are not intrinsic differences between males and females regarding θ, so that we can work with an overall distribution for θ. 31 4.1 The Model for Schooling Each agent chooses the level of schooling, among S possibilities, such that he maximizes his benefit. Let I s represent the net benefit associated with each schooling level s (s={1,…, S }) and assume the following linear-in-the-parameters model for I s : I s = ϕ s Gender + β s X s + α sθ + es for s = 1,K ,S (3) 30 We expect to extend our model to a multi-factor model in which we can precisely distinguish between cognitive and noncognitive abilities. 31 The alternative would have been the estimation of gender specific distributions. We consider this an attractive possibility. However, given the data limitations (sample size) and the large number of parameters in the model, we prefer to follow a simple analysis by considering an overall distribution for θ. Future research should consider the potential differences in unobserved heterogeneity between males and females. 74 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E here ϕ s represents the gender gap associated with the schooling level s, X s is a vector of observed variables determining schooling, β s is the associated vector of parameters, α s is the factor loading associated with the latent ability, and es represents an idiosyncratic component assumed to be independent of θ, and X s . The individual components {e s }s =1 are mutually independent. All of the dependence across schooling choices comes through the observable, X s , and the latent ability θ. S The agent chooses the level of schooling with the highest benefit. Formally, s* = argmax {I s } { s∈ 1,K,S } (4) where s* denotes the individual’s chosen schooling level. Notice that conditional on X s (with s = 1,K ,S ) and θ, equations (3) and (4) can be interpreted as a standard discrete choice model. 4.2 The Model for Accumulated Experience The model also treats accumulated experience as an endogenous outcome. Specifically, after deciding the schooling levels, agents are assumed to pick their experience levels A different alternatives. As in the schooling model, given the schooling level s, we assume a linear-in-the-parameters specification for the benefits associated with the experience level a(s) (I a (s ) ) : I a ( s ) = ϕ a ( s ) Gender + β a ( s ) X a + α a ( s )θ + ea ( s ) for a( s ) = 1,..., A and s=1,…, S where ϕ a (s ) is the gender gap, X a is the vector of observed variables, β a (s ) is the associated vector of parameters, α a (s ) is the factor loading, and ea (s ) represents an components {ea ( s ) }a =1 for any s are mutually independent. Finally, the observed idiosyncratic component assumed to be independent of θ, and X a . The individual A experience level A * ( s*) (where s* represents the schooling level observed in the data) is interpreted as 75 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E A * ( s*) = arg max {I a ( s ) } . a ( s *)∈{1,..., A} 4.3 The Model for Hourly Wages and Monthly Hours Worked For hourly wages and monthly hours worked, we consider schooling/experience specific models. Consider first the model for wages. Denote by s and a(s) the schooling and experience level attained by the individual. Wages ( Ya (s ) ) are modeled using a linear specification: ln Ya ( s ) = ϕY ,a ( s ) Gender + β Y ,a ( s ) X Y + α Y ,a ( s )θ + eY ,a ( s ) for s = 1,K ,S and a( s ) = 1,..., A where ϕ a (s ) is the gender gap, X Y is a vector of observed controls, β Y ,a ( s ) is the vector of coefficients, α a ( s ) is the coefficient associated with the latent ability, and eY ,a ( s ) represents an idiosyncratic error term such that eY ,a ( s ) ⊥ (θ, X Y ) for any a(s)(=1,..., A ) and s(=1,…, S ). A parallel strategy is used to model hours worked. Let H a ( s ) denote the monthly hours worked given schooling level s and experience level a(s). Thus, we assume ln H a ( s ) = ϕ H ,a ( s ) Gender + β H ,a ( s ) X H + α H ,a ( s )θ + ea ( s ) for s = 1,K ,S and a( s ) = 1,..., A where ϕ H ,a ( s ) is the gender gap, X H is a vector of observed controls, β H ,a ( s ) is the vector of coefficients associated with X H , α H ,a ( s ) is the parameters associated with the latent ability, and eH ,a ( s ) represents an idiosyncratic error term such that eH ,a ( s ) ⊥ (θ, X H ) for any a(s)(=1,..., A ) and s(=1,…, S ). 76 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 4.4 The Model for Employment Let I E ,a ( s ) denote the net benefit associated with the alternative of having an employment (versus the alternatives of unemployment or out of the labor force) given the schooling level s and the accumulated experience a(s). As in the previous cases, we assume a linear-in-the-parameters specification for I E ,a ( s ) : I E ,a ( s ) = ϕ E ,a ( s ) Gender + β E ,a ( s ) X E + α E ,a ( s )θ + eE ,a ( s ) for s = 1,K ,S , a( s ) = 1,..., A (5) where ϕ E ,a ( s ) , β E ,a ( s ) , X E , α E ,a ( s ) , and eE ,a ( s ) are defined as before. Finally, the error term is such that eE ,a ( s ) ⊥ (θ, X E ) for any a(s)(=1,..., A ) and s(=1,…, S ). We use (5) to model the employment decisions observed in the data. Specifically, if we let DE ,a ( s ) denote a binary variable such that is equal to 1 if the individual is employed and 0 otherwise, we estimate a binary model assuming that DE ,a ( s ) = 1 I E ,a ( s ) > 0 where 1[.] is (again) the indicator function. [ 4.5 ] Schooling Performance: The Measurement System The identification of the model can be established using the arguments developed in Carneiro, Hansen, and Heckman (2003) and Hansen, Heckman, and Mullen (2004). The identification strategy assumes the existence of a set of measurements. As explained in the next section, these measurements are associated to the individual’s schooling performance. Let T i (i=1,…,n C ) denote the i-th measure. We distinguish the unobserved ability from the observed measure T i . This is important since T i is likely to depend on the characteristics of school as well as on the family background of the individuals by the time of the test. Thus, if X T denote these characteristics, we have Ti = βTi X T + α Ti θ + eTi for i = 1,..., nC where eTi ⊥ (θ , X T ) and eTi ⊥ eT j for any i, j ∈ {1,..., nC } such that i≠j. 77 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Since there are no intrinsic units for the latent ability, we need to normalize one of the loadings in the system to unity to set the scale of the latent ability. Therefore, for some T i (i=1,…,n C ), we set α Ti =1. Notice that our assumptions imply that conditional on observables (variables contained in X), the dependence across all measurements, choices and outcomes come through the unobserved heterogeneity (θ). Notice that if θ were observed, we could use a matching type of approach to control for this dependence (selection). Instead, we estimate the distribution of the unobserved ability and then control for the dependence. Finally, we assume that θ measures the same thing for males and females. 4.6 Implementing the Model The model with unobserved heterogeneity has the following ingredients: the schooling decision problem, the linear models for hourly wages and monthly hours worked (by schooling level s and experience level a(s)), the models for employment (by schooling level s and experience level a(s)), the model for accumulated experience (by schooling level), and finally, the system of measurements or school achievement. Unobserved heterogeneity appears as determinant of each of these components. In this paper we assume that θ is distributed according to a two-component mixture of normals. Formally, ( ) ( ) θ ~ p1 N μ1 , ∑1 + (1 − p1 )N μ1 , ∑ 2 . 2 2 with this assumption we allow a flexible functional form for the unobserved heterogeneity. Following the empirical strategy utilized in Section 3, we estimate the schooling choice model and the experience models using multinomial probit models. Then, we implicitly assume that the idiosyncratic shocks in the equations describing the net utilities are assumed to be jointly normally distributed. The four schooling levels study used here are: primary school, secondary school (or high school), some tertiary education (or some college graduates), and complete tertiary education (or college graduates). For accumulated experience we use three categories: less than 10 years of experience, between 10 and 15 years of experience, and more than 15 years of experience. 78 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E In estimating the model, and since there is no sequential decision process, we use the schooling and experience level reported at the time of the interview. 32 For the models of wages and hours worked we use the information for the month previous to the interview. The same applies in the case of employment status. This is consistent with what we use in Section 3. The measurement system uses the following variables: Average Grade during Secondary Education, Repeated Grade during Primary Education and Repeated Grade during Secondary Education. We normalize the mean of the factor to zero, and we normalize the loading to be equal to one in the equation for the Average Grade during Secondary Education Tables 8A and 8B display the variables used in the empirical implementation of the model, as well as the normalization assuring the identification of the model. The model is estimated using Markov Chain Markov Chain Monte Carlo Methods (MCMC). See Appendix A for a formal discussion of the method used in this paper. 4.7 Main Results Table 9 presents the gender gaps in hourly wages obtained from the model with unobserved heterogeneity. The estimated gaps are in general sizeable and statistically significant. We do not observe clear patterns either by schooling and/or experience levels, although we consistently estimated the largest gender gaps among college graduates (regardless of the experience level considered). In this group we estimate that males make between 36% and 38% more per hour than females. These numbers are larger than those presented in Section 3. But Table 9 also presents a range for the gender gap in wages which goes from -6% (but non significant) for high school dropouts reporting less than 10 years of experience to 38% for college graduates with between 10 and 15 years of experience. Importantly, in only two cases we estimate a gender gap below 15%. 32 In the case of experience, we use the retrospective information provided by the respondent (labor history). The labor history is reported from age 15 or since 1980 depending on the year the individual became 15 years old. For details see Section 2. 79 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Therefore, our evidence indicates the existence of wage differentials that cannot be explained by observed or unobserved characteristics. As in the case of wages, the results obtained for hours worked show a range of values for the gender gaps. These are presented in Table 10. We observe that the point estimates are between -6% (high school dropouts with less than 10 years of experience) and 18% (high school dropouts with between 10 and 15 years of experience). In this case however, less than a half of the estimates are statistically significant. For example, among high school graduates and college graduates we do not find significant gender differences. This is consistent with the evidence presented in Section 3, although the numbers in Table 10 show a broader picture of the gender gaps (if any) in hours worked. Table 11 presents the results for employment. Two are the main results here. First, we observe, in general, a reduction in the estimated gap when we move from low to high experience levels (the only exception is observed among high school graduates). Second, the results suggest that schooling also helps to reduce the estimated gaps (there are only two exceptions in Table 11). In fact, among college graduates the estimated coefficients are -0.12 and -0.23 for experience levels “between 10 and 15 years” and “more than 15 years”, respectively, 33, 34 so the gap favor females in this case. However, as in the case of hours worked, only few estimates are statistically significant, and when significant, they are usually associated with low schooling and experience levels. Table 12 presents the results obtained for the four multinomial choice models used to study accumulated experience. The evidence in Table 12 shows how the gender gap reduces with schooling. Specifically, the significant gender differences estimated for high school dropouts and high school graduates are 100% larger than the ones obtained among individuals with some college. Interestingly, among college graduates we do not find significant differences between genders. Our analysis of gender gaps in variables associated with the labor market leads us to conclude that (1) there are differences between males and females that cannot be explained with observable or unobservable characteristics, and that, in general, (2) these differences are larger among individuals reporting low schooling level and they almost vanish among the more educated individuals. 35 33 For the group of individuals reporting more than 15 years of experience and a college degree, the gender dummy perfectly predicts the labor status: the 29 women in this category reported a job during the week previous to the interview. 34 These coefficients are the point estimates of the parameters associated with the gender dummy variable, so they need to be interpreted cautiously since they do not represent the marginal effects. 35 The only exception to this point, and an important one, comes from the analysis of hourly wages. 80 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E The model also allows us to analyze the gender differences in schooling attainment and schooling achievement. It is worth recalling that the evidence presented in Section 3 already suggested that females outperform males in these two dimensions (see Tables 6 and 7). Table 13 and 14 repeat that analysis but now incorporating unobserved heterogeneity (latent ability). Table 13 presents the gender gaps in schooling decisions. The results show (again) that females are more likely than males to reach higher schooling levels. When compared with those in Table 6, we see that the effects are now larger. Something similar occurs in the case of “repeating a grade during primary school”, “repeating a grade during secondary school”, and “average grades during high school”. The results are shown in Table 14. The evidence in this table suggests that females outperform males, that the differences are statistically significant and that they are larger than the ones presented in Table 7. Specifically, when comparing the estimated gender gaps across tables we obtain 18% (0.26 versus 0.22) and 41% (0.17 versus 0.12) increments in the gender coefficient associated with “repeating a grade in primary school” and “repeating a grade in secondary school”, respectively. In the case of “average grade during secondary school” we obtain an increment of 6.4% in the gender gap (0.33 versus 0.31). 5 Can Unobserved Heterogeneity Explain the Gender Gaps in the Labor Market? From the evidence presented in this paper we must conclude that this is still an open question. Our results do indicate that, after controlling for unobserved heterogeneity, there are non significant gender differences in a variety of labor market variables among educated individuals (e.g., hours worked, accumulated experience, employment), but we still find gender differences among the other schooling groups. These differences can in principle be interpreted as “pure” discrimination. However, this interpretation requires several qualifications. First, our empirical strategy assumes that a one dimensional model of unobserved heterogeneity is sufficient to capture and control for selection (endogeneity) across different margins (decisions). Nevertheless, previous studies have shown the existence of at least two underlying sources of unobserved heterogeneity when explaining labor market outcomes and social behavior. 36 In this context, our one36 See Heckman, Stixrud, and Urzua, 2006; Urzua, 2006; Cunha, Heckman and Navarro, 2004. 81 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E dimensional model might be only partially capturing the unobserved heterogeneity in the data. The consequences of incorporating additional sources of essential heterogeneity on our results are hard to predict. Thus, in principle, we cannot discard the possibility that what we interpret as “unexplained gaps” in the one dimensional case, can be in fact “explained” by, for example, heterogeneity in other unobserved traits (e.g. noncognitive abilities) or preferences (e.g. preferences for leisure). 37 Second, and following up on the previous point, it is interesting to notice that in our results the coefficients associated with what we identify as unobserved heterogeneity are not always significant. The strongest effect of unobserved heterogeneity are obtained for the schooling variables (Tables 13 and 14), and accumulated experience (Table 12). Although the effects are sizeable for the other outcomes, they are usually non-statistically significant. This suggests that our source of unobserved heterogeneity is more closely related to scholastic ability 38 which apparently is not significantly valued in the Chilean labor market after schooling and experience levels are taken into account. Nevertheless, there might be other sources of unobserved heterogeneity that are in fact priced in the labor market. This again illustrates the potential benefits of extending the model to multiple dimensions of unobserved heterogeneity A different consideration regarding the robustness and interpretation of our results can be made by noticing that we implement the model by assuming the existence of a single distribution of unobserved heterogeneity in the sample. The consequences of allowing gender-specific distributions on our previous results are (again) hard to predict, but we believe that the complications of such extension would most likely dominate any potential new insights. This since the identification of gender-specific distribution has additional complications and it relies on even stronger assumptions that the one already made. 39 Besides, from an 37 It is worth noting that the assumption of a single source of unobserved heterogeneity can be relaxed depending on the availability of more comprehensive information at the individual level. These needs for better and more comprehensive information come from the identification argument of the models. Recall that the source of unobserved heterogeneity in this paper is identified using the schooling achievement variable. In order to identify additional sources of heterogeneity we would need additional variables in the measurement system. The availability of information on personality traits, IQ tests, or time preferences could allow the identification and estimation of more general models of unobserved heterogeneity. 38 Before we interpret the source unobserved heterogeneity as a combination of cognitive and noncognitive abilities. 39 Specifically, even though we can assure the identification of gender-specific variance/covariance matrices, the identification of gender-specific mean differences in the distribution of unobserved heterogeneity would require the existence of at least one discrimination free variable. The selection and existence of such variable(s) is non trivial is arguably as well. See Urzua (2006) for details. 82 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E intuitive point of view, we do not find a priori deep reasons to believe that there are gender differences in the distributions of unobserved heterogeneity. It is because of this remarks that the estimation of gender-specific distribution is left for future research. 6 Conclusions In this paper we present a comprehensive analysis of the gender gaps in a variety of labor market outcomes for Chile. The analysis is carried out using two different approaches. The first approach follows the literature by estimating linear and nonlinear models of a variety of variables on different observable controls and the gender dummy. This approach does not pay attention to potential selection problems (endogeneity). The second approach is more general. It allows for the presence of individuals’ unobserved heterogeneity that is assumed to be the cause of the endogeneity problems in the conventional approach. Our main results are robust across the approaches. They suggest the existence of gender gaps in labor market variables that cannot be explained by observable or unobservable characteristics or by underlying selection mechanisms generating endogeneity. Nevertheless, the findings from the model with unobserved heterogeneity indicate that the gender gaps critically depend on the schooling level of the individuals considered in the analysis. This is particular important among college graduates. For this group, the gender differences are in general non-significant. The evidence also demonstrates that females outperform males in schooling achievement and schooling performance. This is observed regardless of the approach, but we find the stronger effects in the model with unobserved heterogeneity. These gender differences favoring women represent an argument against the conventional idea that labor market differences can be interpreted as the result of differences in the human capital between genders. Obviously, this conclusion assumes that the utilized variables are good proxies of the actual human capital accumulated by the individuals. Overall, the estimates in this paper could lead us to conclude that women are effectively discriminated in the labor market with the largest gender gaps observed among the less educated people. However, we prefer to interpret our results cautiously. We believe that the availability of better data and the estimation of 83 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E even more general models than the one considered here could indeed explain some of the unexplained estimated gender gaps. 84 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E References Carneiro, P., K. Hansen, and J. Heckman (2003). Estimating distributions of treatment effects with an application to the returns to schooling and measurement of the effects of uncertainty on college choice. International Economic Review 44 (2), 361—422. 2001 Lawrence R. Klein Lecture. Cunha, F., J. Heckman, and S. Navarro. Separating Uncertainty from Heterogeneity in Life Cycle Earnings. The 2004 Hicks Lecture, Oxford Economic Papers 57(191-261), 1-72 Hansen, K., J. Heckman, and K. Mullen (2004). The effect of schooling and ability on achievement test scores. Journal of Econometrics 121 (1-2), 39—98. Heckman, J. (1974). Shadow prices, market wages, and labor supply. Econometrica 42(4), 679. Heckman, J. (1981). Statistical models for discrete panel data. In C. Manski and D. Mc-Fadden (Eds.), Structural Analysis of Discrete Data with Econometric Applications, pp. 114—178. Cambridge, MA: MIT Press. Heckman, J., J. Stixrud, and S. Urzua (2006). The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. Journal of Labor Economics, 24(3). Urzua, S. (2006). Racial Labor Market Gaps: The Role of Abilities and Schooling Choices. Working paper, University of Chicago. 85 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 86 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 87 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 88 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 89 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 90 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 91 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 92 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E 93 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 1. Means of Schooling and Labor Market Outcomes by Gender SPS02 Females Variable (Dummy=1 if Apply) Mean Std. Dev A. School Information Maximum Schooling Level = Primary Education 0.11 0.32 Maximum Schooling Level = Secondary Education 0.51 0.50 Maximum Schooling Level = Some Tertiary Education 0.26 0.44 Maximum Schooling Level = Complete Tertiary Education 0.11 0.31 Repeat a Grade in Primary School 0.22 0.41 Repeat a Grade in Secondary School 0.20 0.40 (a) Average Grade in Secondary School 0.16 0.98 Mean 0.17 0.49 0.24 0.10 0.30 0.24 -0.17 Males Std. Dev 0.38 0.50 0.43 0.30 0.46 0.43 1.00 B. Labor Market Variables Monthly Earnings Hours Worked per Week Hourly Wage Working During Last Month Less than 10 years of Experience Between 10 and 15 years of Experience More than 15 years of Experience 215,266 43.41 1,292 0.59 0.56 0.26 0.18 214,323 285,140 360,046 11.74 48.17 9.81 1,257 1,636 4,649 0.49 0.82 0.39 0.50 0.25 0.43 0.44 0.34 0.47 0.39 0.41 0.49 1,765 1,801 Number of Observations Note: The numbers presented in this table corresponds to the sample of individuals with ages between 28 and 40 years old at the time of the interview. 94 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 2. The Gender Gap in Hourly Wages SPS02 Variables Male Schooling (B) 0.23 (0.03) (C) 0.23 (0.03) (D) 0.29 (0.03) - - 0.29 (0.04) 0.49 (0.04) 0.90 (0.06) 0.30 (0.04) 0.50 (0.05) 0.92 (0.06) - 0.04 (0.03) 0.04 (0.03) 0.05 (0.03) 0.10 (0.03) 0.14 (0.03) 0.19 (0.04) -0.15 (0.04) -0.04 (0.04) 0.22 (0.03) -0.15 (0.04) -0.04 (0.04) 0.22 (0.03) -0.15 (0.04) -0.05 (0.04) 0.21 (0.03) -0.15 (0.04) -0.004 (0.04) 0.24 (0.03) -0.13 (0.03) -0.08 (0.07) -0.13 (0.03) -0.08 (0.07) -0.10 (0.03) -0.04 (0.07) -0.11 (0.03) -0.06 (0.07) 0.09 (0.07) -0.33 (0.07) -0.71 (0.07) -1.08 (0.07) -1.35 (0.08) -1.05 (0.06) -1.11 (0.07) -1.28 (0.07) 7.63 (0.07) No 0.10 (0.07) -0.33 (0.07) -0.72 (0.07) -1.08 (0.07) -1.36 (0.08) -1.05 (0.06) -1.11 (0.07) -1.28 (0.07) 7.61 (0.07) No -0.18 (0.07) -0.27 (0.07) -0.56 (0.06) -0.84 (0.07) -0.96 (0.08) -0.77 (0.06) -0.85 (0.07) -0.94 (0.07) 7.04 (0.08) No -0.17 (0.07) -0.25 (0.07) -0.53 (0.06) -0.83 (0.07) -0.93 (0.09) -0.74 (0.06) -0.82 (0.07) -0.91 (0.07) 6.75 (0.10) Yes (a) Secondary Education Some Tertiary Education Complete Tertiary Education Experience (b) Between 10 and 15 years of Experience More than 10 years of Experience Residence (A) 0.24 (0.03) (c) Central South Santiago Type of Job (d) Employer or Self-Worker Domestic Service Occupations (e) Professionals Technicians and associate professionals Clerks Service workers and shop and market sales workers Skilled agricultural and fishery workers Craft and related trades workers Plant and machine operators and assemblers Elementary occupations Constant Correction for Selection Notes: (a) The baseline category is Primary Education; (b) The baseline category is Less than 10 years of experience; (c) The baseline category is North (I to III regions). Central represents IV-VII regions (including the XIII region), South represents VIII-XII regions; (d) The baseline category is Public and Private Employees; (e) The baseline category is Legislators, senior officials and managers. For each model Shooling corresponds to the declared schooling level for each individual in the sample. Specification (D) includes the same controls as (C) but is estimated including a correction for selection. The variables used in the first stage are number of children, mother's occupational situation, father's occupational situation, and whether or not the individual grew up in a poor household. Standard Errors are presented in parentheses. 95 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 3. The Gender Gap in Monthly Hours Worked SPS02 Variables Male Schooling (B) 0.11 (0.02) (C) 0.11 (0.02) (D) 0.004 (0.02) - - -0.01 (0.02) 0.02 (0.03) -0.03 (0.04) -0.04 (0.02) -0.03 (0.02) -0.04 (0.03) - 0.08 (0.02) 0.08 (0.02) 0.08 (0.02) 0.08 (0.02) -0.07 (0.02) -0.08 (0.02) -0.002 (0.03) -0.05 (0.02) 0.02 (0.02) -0.005 (0.03) -0.05 (0.02) 0.02 (0.02) -0.01 (0.03) -0.05 (0.02) 0.02 (0.02) 0.02 (0.03) -0.10 (0.03) -0.03 (0.02) -0.20 (0.02) -0.11 (0.04) -0.20 (0.02) -0.12 (0.04) -0.20 (0.02) -0.12 (0.04) -0.05 (0.02) 0.01 (0.03) -0.30 (0.04) -0.24 (0.04) -0.18 (0.04) -0.19 (0.04) -0.18 (0.05) -0.16 (0.04) -0.12 (0.04) -0.24 (0.04) 3.95 (0.04) No -0.28 (0.04) -0.24 (0.04) -0.18 (0.04) -0.19 (0.04) -0.20 (0.05) -0.17 (0.04) -0.13 (0.04) -0.25 (0.04) 3.91 (0.04) No -0.27 (0.04) -0.25 (0.04) -0.19 (0.04) -0.19 (0.04) -0.20 (0.05) -0.18 (0.04) -0.13 (0.04) -0.25 (0.04) 3.92 (0.05) No -0.22 (0.03) -0.17 (0.03) -0.16 (0.03) -0.11 (0.03) -0.16 (0.04) -0.13 (0.03) -0.06 (0.03) -0.17 (0.03) 4.21 (0.04) Yes (a) Secondary Education Some Tertiary Education Complete Tertiary Education Experience (b) Between 10 and 15 years of Experience More than 10 years of Experience Residence (A) 0.12 (0.02) (c) Central South Santiago Type of Job (d) Employer or Self-Worker Domestic Service Occupations (e) Professionals Technicians and associate professionals Clerks Service workers and shop and market sales workers Skilled agricultural and fishery workers Craft and related trades workers Plant and machine operators and assemblers Elementary occupations Constant Correction for Selection Notes: (a) The baseline category is Primary Education; (b) The baseline category is Less than 10 years of experience; (c) The baseline category is North (I to III regions). Central represents IV-VII regions (including the XIII region), South represents VIII-XII regions; (d) The baseline category is Public and Private Employees; (e) The baseline category is Legislators, senior officials and managers. For each model Shooling corresponds to the declared schooling level for each individual in the sample. Specification (D) includes the same controls as (C) but is estimated including a correction for selection. The variables used in the first stage are number of children, mother's occupational situation, father's occupational situation, and whether or not the individual grew up in a poor household. Standard Errors are presented in parentheses. 96 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 4. The Gender Gap in Employment SPS02 Variables Coefficient 0.67 (0.05) Male (A) Marg. Effect 0.22 (0.02) Coefficient 0.42 (0.05) (B) Marg. Effect 0.14 (0.02) Coefficient 0.41 (0.05) (C) Marg. Effect 0.14 (0.02) Background (a) Number of Children Age Mother's Occupation Father's Occupation Growing Up in Poverty Schooling -0.03 (0.01) 0.01 (0.00) -0.02 (0.02) -0.08 (0.08) -0.08 (0.02) -0.08 (0.02) -0.03 (0.01) -0.03 (0.05) -0.21 (0.30) -0.27 (0.05) -0.03 (0.01) -0.01 (0.00) -0.01 (0.02) -0.07 (0.09) -0.09 (0.02) -0.04 (0.02) -0.04 (0.01) -0.06 (0.05) -0.13 (0.31) -0.14 (0.05) -0.01 (0.01) -0.01 (0.00) -0.02 (0.02) -0.04 (0.09) -0.05 (0.02) - - - - 0.26 (0.07) 0.59 (0.08) 1.22 (0.12) 0.09 (0.02) 0.17 (0.02) 0.27 (0.01) - - 0.66 (0.06) 0.88 (0.07) 0.20 (0.02) 0.26 (0.02) 0.73 (0.06) 1.04 (0.08) 0.21 (0.02) 0.29 (0.02) -0.08 (0.08) 0.22 (0.08) 0.24 (0.06) 0.03 (0.37) -0.03 (0.03) 0.07 (0.03) 0.08 (0.02) - -0.09 (0.08) 0.24 (0.08) 0.24 (0.06) 1.25 (0.40) -0.03 (0.03) 0.08 (0.03) 0.08 (0.02) - -0.06 (0.08) 0.24 (0.08) 0.18 (0.06) 0.91 (0.42) -0.02 (0.03) 0.08 (0.03) 0.06 (0.02) - (b) Secondary Education Some Tertiary Education Complete Tertiary Education Experience (c) Between 10 and 15 years of Experience More than 10 years of Experience Residence -0.09 (0.02) 0.02 (0.01) -0.05 (0.05) -0.27 (0.29) -0.24 (0.05) (d) Central South Santiago Constant Notes: (a) Mother's and Father's Education are dummy variables that take a value of one if the respective parent worked as asalaried and zero otherwise; (b) The baseline category is Primary Education; (c) The baseline category is Less than 10 years of experience; (d) The baseline category is North (I to III regions). Central represents IVVII regions (including the XIII region), South represents VIII-XII regions. Standard Errors are presented in parentheses. 97 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 5. The Gender Gap in Accumulated Experience SPS02 Variables (a) (b) Male Secondary Education Some College College Graduates Mother's Years of Schooling Father's Years of Schooling Growing Up in Poverty Growing Up in Broken Home Age Constant Less Than 10 Years Coefficient Marg. Effect 1.11 -0.40 (0.07) (0.02) 0.26 -0.04 (0.11) (0.03) 0.08 0.04 (0.13) (0.04) -0.07 0.11 (0.16) (0.04) -0.01 0.002 (0.01) (0.003) -0.02 0.01 (0.01) (0.003) -0.05 0.003 (0.08) (0.02) -0.15 0.01 (0.17) (0.05) 0.11 -0.07 (0.01) (0.00) -4.10 (0.40) (a) Between 10 and 15 Years Coefficient Marg. Effect 1.92 0.11 (0.09) (0.02) -0.08 0.09 (0.12) (0.03) -0.61 0.08 (0.14) (0.03) -1.16 0.07 (0.19) (0.04) 0.00 -0.002 (0.01) (0.003) -0.04 -0.003 (0.01) (0.003) 0.06 -0.02 (0.09) (0.02) 0.16 -0.06 (0.21) (0.04) 0.42 -0.01 (0.01) (0.00) -15.07 (0.55) More than 15 Years Marg. Effect 0.29 (0.02) -0.04 (0.02) -0.13 (0.02) -0.18 (0.02) -0.0003 (0.003) -0.01 (0.003) 0.02 (0.02) 0.05 (0.03) 0.08 (0.00) (a) Notes: (a) The experience levels correspond to the accumulated experience declared during the interview. Post-secondary education includes includes technical education (complete and incomplete). (b) The shooling level corresponds to the schooling level declared in the sample. Post-secondary education includes includes technical education (complete and incomplete). Table 6. The Gender Gap in Schooling Decisions SPS02 Variables Male Mother's Years of Schooling Father's Years of Schooling Growing Up in Poverty Growing Up in Broken Home Age Constant Primary School Coefficient Marg. Effect -0.30 0.04 (0.08) (0.01) 0.08 -0.02 (0.02) (0.00) 0.05 -0.01 (0.01) (0.00) -0.59 0.11 (0.08) (0.01) 0.43 -0.09 (0.17) (0.03) -0.03 0.004 (0.01) (0.001) 1.07 (0.40) Secondary School Coefficient Marg. Effect -0.33 -0.02 (0.09) (0.02) 0.15 -0.01 (0.02) (0.00) 0.13 -0.02 (0.01) (0.00) -0.84 0.00 (0.09) (0.02) 0.83 0.00 (0.21) (0.04) -0.04 -0.002 (0.01) (0.002) -0.68 (0.45) Some Post-Secondary Education College Graduates Coefficient Marg. Effect Marg. Effect -0.30 -0.02 -0.004 (0.10) (0.02) (0.01) 0.17 0.02 0.01 (0.02) (0.00) (0.002) 0.16 0.02 0.01 (0.02) (0.00) (0.002) -0.84 -0.08 -0.03 (0.11) (0.02) (0.01) 0.34 0.11 -0.01 (0.23) (0.03) (0.03) -0.01 -0.004 0.003 (0.01) (0.002) (0.001) -2.35 (0.52) Notes: The shooling level corresponds to the schooling level declared in the sample. Post-secondary education includes includes technical education (complete and incomplete). 98 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 7. The Gender Gap in Schooling Achievement SPS02 Variables Repeating a Grade in Primary School Coefficient Marg. Effect 0.22 0.07 (0.05) (0.01) Male Mother's Education Average Score during (a) Secondary School Coefficient -0.31 (0.04) (b) Secondary Education Some Tertiary Education Complete Tertiary Education Father's Education Repeating a Grade in Secondary School Coefficient Marg. Effect 0.12 0.04 (0.05) (0.02) -0.06 (0.06) 0.14 (0.19) -0.29 (0.22) -0.02 (0.02) 0.04 (0.06) -0.08 (0.06) 0.02 (0.06) -0.06 (0.20) -0.13 (0.20) 0.01 (0.02) -0.02 (0.06) -0.04 (0.05) 0.05 (0.04) 0.11 (0.13) 0.35 (0.13) -0.17 (0.06) -0.51 (0.16) -0.41 (0.16) -0.05 (0.02) -0.13 (0.03) -0.11 (0.04) -0.05 (0.06) -0.29 (0.16) -0.11 (0.15) -0.01 (0.02) -0.08 (0.04) -0.03 (0.04) 0.14 (0.04) 0.23 (0.10) 0.21 (0.10) 0.25 (0.05) -0.38 (0.11) 0.08 (0.02) -0.13 (0.04) -0.04 (0.06) 0.04 (0.13) -0.01 (0.02) 0.01 (0.04) -0.16 (0.04) 0.10 (0.09) (b) Secondary Education Some Tertiary Education Complete Tertiary Education Background Growing Up in Poverty Growing Up in Broken Home School Characteristics (c) Urban Primary School -0.20 -0.07 0.02 (0.08) (0.03) (0.08) Urban Secondary School 0.40 0.10 0.22 (0.24) (0.05) (0.16) Private-Subsized Primary School -0.10 -0.03 0.07 (0.07) (0.02) (0.06) Coorporation - Primary School -0.45 -0.12 0.22 (0.59) (0.12) (0.35) Private Primary Schoo -0.27 -0.08 0.09 (0.12) (0.03) (0.09) Private-Subsized Secondary School -0.21 -0.06 0.13 (0.06) (0.02) (0.05) Coorporation - Secondary School -0.42 -0.10 0.15 (0.26) (0.05) (0.17) Private Secondary School -0.41 -0.10 0.24 (0.12) (0.03) (0.10) Constant -0.18 -1.15 -0.33 (0.13) (0.27) (0.17) Notes: (a) The average score is standarized to have mean 0 and variance 1 in the population; (b) The baseline category is Primary Education; (c) In the case of the dummies controlling for the type of management the baseline category is Public School. Standard Errors are presented in parentheses. 99 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 8A. Variables in the empirical implementation of the model Outcome Equations Hourly Wage (a) Monthly Hours Accumulated Educational Choice Employment (a) Worked (a) Experience (b) Model (c) Gender Dummy Yes Yes Yes Yes Yes Region of Residence Yes Yes Yes Growing Up in Broken Home Yes Mother's Education Yes Yes Father's Education Yes Yes Growing Up in Poverty Yes Yes Age Yes Yes Yes Type of Occupation Yes Yes Type of Job Yes Yes Unobserved Ability Yes Yes Yes Yes Yes Notes: (a) Hourly wages, monthly hours worked and employment models are estimated for four different schooling categories (primary, secondary, some tertiary and complete tertiary) and three different levels of accumulated experience (less than 10 years, between 10 and 15 years, and more than 15 years). In each case, the labor market outcome refers to the previous month individual's outcome; (b) Accumulated experience is modeled with a multinomial choice model. The categories considered are: less than 10 years, between 10 and 15 years, and more than 15 years. The level of accumulated experience is the total work experience reported at the time of the interview; Variables Table 8B. Variables in the empirical implementation of the model Auxiliary Measures Variables Primary School in a Urban Area (Dummy) Secondary School in a Urban Area (Dummy) Growing Up in Broken Home Mother's Education Father's Education Growing Up in Poverty Primary School System (Public, Private, etc.) Secondary School System (Public, Private, etc.) Unobserved Ability Average Grade in Repeat Any Secondary Grade in Education Primary School Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Repeat Any Grade in Secondary School Yes Yes Yes Yes Yes Yes 1.0 100 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 9. Model with Essential Heterogeneity Gender Gap in Hourly Wages, by Schooling Level and Accumulated Experience (a) SPS02 High School Dropouts High School Graduates Some Post-Secondary Education College Graduates Less than Between 10 More than 15 Less than Between 10 More than 15 Less than 10 Between 10 More than 15 Less than Between 10 More than 15 and 15 Years 10 Years 10 Years and 15 Years Years and 15 Years Years 10 Years and 15 Years Male -0.06 0.30 0.07 0.35 0.15 0.19 0.23 0.35 0.15 0.38 0.38 0.36 (0.29) (0.18) (0.14) (0.09) (0.06) (0.05) (0.08) (0.08) (0.10) (0.08) (0.15) (0.20) Employer or Self-Worker (b) -0.41 -0.34 -0.30 0.19 -0.19 -0.23 0.22 0.06 -0.12 0.00 -0.10 0.41 (0.37) (0.15) (0.10) (0.11) (0.08) (0.06) (0.12) (0.14) (0.14) (0.15) (0.20) (0.39) Domestic Service -0.52 0.18 -0.27 -0.11 -0.13 0.16 0.08 -1.44 -0.11 (0.37) (0.24) (0.20) (0.16) (0.17) (0.17) (0.28) (0.61) (0.50) (c) Professionals -1.04 -0.52 0.26 0.42 -0.11 -0.21 -0.18 -0.18 (0.63) (0.48) (0.22) (0.36) (0.29) (0.14) (0.24) (0.30) Technicians and associate professionals -0.96 -0.36 -0.40 0.03 0.26 -0.41 -0.11 -0.46 -0.03 (0.28) (0.20) (0.15) (0.18) (0.25) (0.23) (0.18) (0.30) (0.35) Clerks -0.83 -1.22 -0.48 -0.43 -0.38 -0.11 -0.53 -0.55 -0.79 -0.56 (0.60) (0.25) (0.18) (0.13) (0.19) (0.26) (0.24) (0.19) (0.39) (0.47) Service workers and shop and market sales workers -0.57 -0.48 -1.66 -0.61 -0.84 -0.64 -0.46 -0.59 -0.84 (0.48) (0.32) (0.25) (0.18) (0.13) (0.20) (0.26) (0.24) (0.52) Skilled agricultural and fishery workers 0.39 -0.55 -0.78 -1.75 -0.83 -0.92 -0.44 0.47 -0.81 (0.57) (0.40) (0.29) (0.37) (0.24) (0.16) (0.63) (0.46) (0.38) Craft and related trades workers 0.25 -0.38 -0.57 -1.44 -0.68 -0.67 -0.57 -0.27 -0.63 -1.22 (0.34) (0.38) (0.28) (0.26) (0.17) (0.12) (0.22) (0.26) (0.25) (0.85) Plant and machine operators and assemblers 0.69 -0.36 -0.57 -1.52 -0.72 -0.81 -0.80 -0.52 -1.11 0.59 -1.37 (0.56) (0.40) (0.30) (0.26) (0.17) (0.12) (0.27) (0.27) (0.32) (0.69) (0.59) Elementary occupations 0.35 -0.67 -0.63 -1.66 -0.80 -0.88 -0.91 -0.55 -1.18 -0.97 -1.75 (0.32) (0.38) (0.28) (0.26) (0.20) (0.13) (0.28) (0.30) (0.28) (0.38) (0.50) Central 0.48 -0.11 -0.13 -0.29 -0.21 -0.15 -0.05 -0.19 0.11 -0.13 0.38 -0.58 (0.51) (0.22) (0.13) (0.12) (0.09) (0.07) (0.15) (0.15) (0.23) (0.17) (0.28) (0.38) South 0.41 -0.13 -0.03 -0.24 -0.07 0.03 -0.10 -0.13 0.11 0.03 0.42 -0.12 (0.43) (0.21) (0.14) (0.11) (0.09) (0.08) (0.13) (0.14) (0.23) (0.15) (0.27) (0.34) Santiago -0.25 0.36 0.19 0.22 0.25 0.27 0.08 0.20 0.12 0.35 -0.05 0.39 (0.35) (0.16) (0.11) (0.10) (0.07) (0.06) (0.11) (0.10) (0.13) (0.11) (0.19) (0.26) Intercept 5.56 7.04 7.04 8.10 7.36 7.37 7.28 7.12 7.52 8.07 7.98 8.80 (0.47) (0.51) (0.34) (0.27) (0.19) (0.13) (0.21) (0.30) (0.30) (0.25) (0.43) (0.55) Unobserved Heterogeneity -0.20 0.71 -0.03 0.17 0.13 -0.19 -0.32 -0.30 -0.004 -0.39 -0.50 -0.89 (0.13) (0.43) (0.07) (0.15) (0.11) (0.09) (0.15) (0.17) (0.20) (0.22) (0.39) (0.51) Notes: The accumulated experience corresponds to the retrospective information reported by the individual at the time of the interview. The shooling level corresponds to the schooling level declared in the sample. Postsecondary education includes includes technical education (complete and incomplete). (b) For the characteristics of the type of job (employer or self-worker and domestic service), the baseline category is Public and Private Employees; (c) For the set of variables controlling for occupation characteristics (from Professionals to Elementary Occupations in this table) the baseline category is Legislators, senior officials and managers. Variable Table 10. Model with Essential Heterogeneity Gender Gap in Hours Worked, by Schooling Level and Accumulated Experience (a) SPS02 High School Dropouts High School Graduates Some Post-Secondary Education College Graduates Less than Between 10 More than Less than 10 Between 10 More than Less than 10 Between 10 More than 15 Less than Between 10 More than and 15 15 Years Years 10 Years and 15 Years 15 Years Years and 15 Years Years 10 Years and 15 15 Years Male -0.06 0.18 0.14 0.08 0.10 0.07 0.17 0.12 0.10 0.02 0.00 0.08 (0.15) (0.13) (0.08) (0.07) (0.03) (0.03) (0.05) (0.04) (0.05) (0.06) (0.08) (0.10) -0.01 -0.12 -0.17 -0.53 -0.06 -0.16 -0.24 -0.31 -0.22 -0.09 -0.02 -0.07 Employer or Self-Worker (b) (0.18) (0.11) (0.06) (0.09) (0.04) (0.03) (0.08) (0.07) (0.07) (0.11) (0.10) (0.20) Domestic Service 0.18 -0.09 0.08 -0.25 0.00 -0.18 0.07 -0.75 -0.09 (0.19) (0.18) (0.11) (0.14) (0.08) (0.09) (0.14) (0.30) (0.36) Professionals (c) -0.96 -0.08 -0.34 -0.45 -0.33 -0.20 -0.29 -0.15 (0.55) (0.26) (0.15) (0.18) (0.15) (0.10) (0.13) (0.15) Technicians and associate professionals -0.34 -0.05 -0.21 -0.42 -0.38 -0.22 -0.19 -0.19 -0.31 (0.24) (0.10) (0.08) (0.12) (0.12) (0.11) (0.13) (0.15) (0.17) Clerks 0.31 -0.21 -0.11 -0.25 -0.27 -0.34 -0.22 -0.02 -0.15 -0.02 (0.34) (0.22) (0.09) (0.07) (0.13) (0.13) (0.12) (0.14) (0.19) (0.24) Service workers and shop and market sales workers 0.36 -0.27 -0.11 -0.11 -0.23 -0.38 -0.31 -0.28 -0.02 0.00 (0.36) (0.20) (0.21) (0.09) (0.07) (0.13) (0.13) (0.12) (0.28) (0.27) Skilled agricultural and fishery workers 0.01 -0.10 -0.09 -0.31 -0.06 -0.22 0.11 -0.29 -0.09 (0.29) (0.29) (0.19) (0.30) (0.12) (0.09) (0.41) (0.23) (0.18) Craft and related trades workers -0.11 0.07 -0.06 -0.23 -0.05 -0.23 -0.27 -0.33 -0.09 0.38 (0.17) (0.28) (0.18) (0.22) (0.08) (0.06) (0.15) (0.13) (0.12) (0.43) Plant and machine operators and assemblers 0.06 0.03 -0.08 -0.09 -0.02 -0.16 -0.06 -0.25 -0.43 -0.06 0.02 (0.28) (0.30) (0.20) (0.23) (0.09) (0.06) (0.18) (0.13) (0.16) (0.36) (0.29) Elementary occupations -0.16 0.01 -0.09 -0.28 -0.09 -0.27 -0.68 -0.64 -0.05 -0.52 (0.16) (0.28) (0.18) (0.22) (0.10) (0.07) (0.18) (0.15) (0.14) (0.26) Central 0.05 -0.07 0.05 -0.04 -0.02 -0.02 0.13 -0.05 -0.12 -0.07 -0.26 -0.20 (0.25) (0.16) (0.08) (0.10) (0.05) (0.04) (0.09) (0.07) (0.12) (0.12) (0.14) (0.20) South -0.19 -0.01 -0.15 0.00 0.00 -0.04 0.07 -0.04 -0.07 -0.19 -0.22 -0.31 (0.22) (0.16) (0.08) (0.09) (0.04) (0.04) (0.09) (0.07) (0.12) (0.11) (0.14) (0.17) Santiago -0.09 -0.02 -0.02 0.07 0.05 0.01 -0.03 0.03 0.09 0.02 0.02 0.03 (0.17) (0.12) (0.06) (0.09) (0.04) (0.03) (0.07) (0.05) (0.07) (0.08) (0.10) (0.14) Intercept 4.35 3.54 4.01 3.91 3.85 4.05 3.92 4.17 4.06 3.90 4.04 4.05 (0.27) (0.37) (0.22) (0.22) (0.09) (0.07) (0.14) (0.14) (0.15) (0.17) (0.21) (0.27) Unobserved Heterogeneity 0.40 -0.39 0.23 -0.15 0.01 0.03 0.18 0.06 0.08 0.08 0.17 -0.04 (0.13) (0.26) (0.06) (0.12) (0.05) (0.05) (0.10) (0.08) (0.10) (0.13) (0.17) (0.22) Notes: The accumulated experience corresponds to the retrospective information reported by the individual at the time of the interview. The shooling level corresponds to the schooling level declared in the sample. Post-secondary education includes includes technical education (complete and incomplete). (b) For the characteristics of the type of job (employer or self-worker and domestic service), the baseline category is Public and Private Employees; (c) For the set of variables controlling for occupation characteristics (from Professionals to Elementary Occupations in this table) the baseline category is Legislators, senior officials and managers. Variable 1 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 11. Model with Essential Heterogeneity Gender Gap in Employment Status, by Schooling Level and Accumulated Experience SPS02 (a) (b) High School Dropouts High School Graduates College Graduates Some Post-Secondary Education Less than Between 10 More than Less than Between 10 More than Less than Between 10 More than Less than 10 Between 10 10 Years and 15 15 Years 10 Years and 15 15 Years 10 Years and 15 15 Years Years and 15 Years Male 1.40 0.35 -0.10 0.98 0.30 0.34 0.80 0.36 0.18 -0.12 -0.23 (0.54) (0.35) (0.35) (0.14) (0.12) (0.13) (0.14) (0.20) (0.24) (0.25) (0.47) Central -0.17 -0.09 0.00 0.03 0.42 0.42 -0.39 (0.17) (0.21) (0.22) (0.23) (0.32) (0.48) (0.61) South 0.16 0.05 -0.06 0.29 1.17 1.45 -0.36 (0.17) (0.20) (0.22) (0.23) (0.37) (0.62) (0.59) Santiago 0.17 0.27 0.10 -0.06 0.20 0.10 -0.19 (0.14) (0.16) (0.16) (0.18) (0.26) (0.33) (0.33) Number of Children 0.09 0.04 0.11 -0.11 -0.07 -0.07 -0.04 -0.03 0.01 0.14 0.17 (0.10) (0.13) (0.10) (0.05) (0.05) (0.05) (0.07) (0.09) (0.10) (0.13) (0.21) Intercept -2.20 -0.26 0.13 -0.08 0.62 0.79 0.06 0.40 0.38 1.13 3.16 (1.17) (0.54) (0.61) (0.17) (0.20) (0.23) (0.21) (0.29) (0.44) (0.72) (2.09) Unobserved Heterogeneity -1.65 -1.63 -1.64 0.37 0.21 0.10 0.07 0.07 0.22 0.59 -1.69 (1.35) (1.27) (1.62) (0.22) (0.27) (0.25) (0.29) (0.44) (0.54) (0.70) (1.81) Notes: The accumulated experience corresponds to the retrospective information reported by the individual at the time of the interview. The shooling level corresponds to the schooling level declared in the sample. Post-secondary education includes includes technical education (complete and incomplete). (a) Among high school dropouts, the characteristics of the place of residence perfectly predict the labor status, so those variables are excluded in these cases. (b) For the group of individuals reporting more than 15 years of experience and a college degree, the gender dummy perfectly predicts the labor status: the 29 women in these category reported to be working (34 out of 37 males report to be working). Since the gender coefficient is the main interest of this table we do not include this model here. Variable 2 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 12. Model with Essential Heterogeneity Gender Gap in Accumulated Experience, by Schooling Level SPS02 High School Dropouts High School Graduates Some College College Graduates Between 10 More than Between 10 More than 15 Between 10 More than 15 Between 10 More than and 15 Years 15 Years and 15 Years Years and 15 Years 15 Years and 15 Years Years Male 1.47 2.95 1.48 2.50 0.75 1.02 0.14 0.41 (0.23) (0.24) (0.12) (0.15) (0.14) (0.21) (0.28) (0.42) Mother's Years of Schooling 0.04 -0.02 0.03 0.03 -0.01 0.05 -0.04 -0.11 (0.05) (0.05) (0.02) (0.02) (0.02) (0.04) (0.06) (0.09) Father's Years of Schooling -0.04 0.00 0.00 -0.02 -0.02 0.00 0.09 -0.06 (0.05) (0.04) (0.02) (0.02) (0.02) (0.04) (0.06) (0.10) Growing Up in Poverty -0.21 -0.03 -0.16 -0.14 0.02 0.25 -0.30 0.79 (0.24) (0.22) (0.12) (0.14) (0.19) (0.26) (0.42) (0.69) Age -0.05 0.26 0.11 0.46 0.20 0.61 0.39 0.78 (0.03) (0.03) (0.02) (0.03) (0.03) (0.06) (0.06) (0.11) Intercept 1.37 -10.45 -4.22 -17.06 -6.98 -22.97 -16.69 -25.84 (1.29) (1.21) (0.64) (0.97) (1.00) (2.73) (3.54) (4.42) Unobserved Heterogeneity 1.19 -0.20 1.23 1.57 0.53 2.09 2.95 -1.71 (0.74) (0.17) (0.28) (0.36) (0.45) (0.96) (1.55) (3.10) Notes: The accumulated experience corresponds to the retrospective information reported by the individual at the time of the interview. The table presents the results for three multinomial choice models (each for each schooling level). The baseline category is less than 10 years of accumulated experience. Variables 3 D EPA RT A M EN T O D E EC O N O M Í A , U N I V ERS I D A D D E C H I L E Table 13. Model with Essential Heterogeneity Gender Gap in Schooling Decisions SPS02 Variable Secondary Some PostCollege School Secondary Graduates Male -0.47 -0.55 -0.61 (0.11) (0.13) (0.30) Mother's Years of Schooling 0.13 0.23 0.41 (0.02) (0.03) (0.06) Father's Years of Schooling 0.09 0.21 0.44 (0.02) (0.02) (0.07) Growing Up in Poverty -0.03 -0.03 0.08 (0.01) (0.02) (0.04) Growing Up in Broken Home 0.53 1.02 0.46 (0.22) (0.30) (0.71) Age -0.81 -1.25 -2.20 (0.11) (0.15) (0.46) Intercept 1.10 -1.66 -12.93 (0.51) (0.64) (2.99) Unobserved Heterogeneity 1.90 3.52 10.90 (0.38) (0.48) (1.96) Notes: The shooling level corresponds to the schooling level declared in the sample. Post-secondary education includes includes technical education (complete and incomplete). The baseline category is Primary School. Table 14. Model with Essential Heterogeneity Gender Gap in Hours Schooling Achievement SPS02 Variables (a) Repeating a Grade in Primary School Repeating a Grade in Secondary School Average Score during Secondary School (b) Male 0.26 0.17 -0.33 (0.05) (0.06) (0.04) Mother: Secondary Education -0.08 -0.04 0.10 (0.07) (0.07) (0.04) Mother: Some Tertiary Education 0.06 -0.20 0.23 (0.21) (0.22) (0.12) Mother: Complete Tertiary Education -0.30 -0.18 0.38 (0.25) (0.23) (0.12) Father: Secondary Education -0.20 -0.12 0.20 (0.06) (0.07) (0.04) Father: Some Tertiary Education -0.61 -0.44 0.30 (0.18) (0.18) (0.10) Father: Complete Tertiary Education -0.40 -0.17 0.24 (0.17) (0.17) (0.09) Growing Up in Poverty 0.28 0.04 -0.23 (0.06) (0.07) (0.04) Growing Up in Broken Home -0.09 0.45 0.01 (0.08) (0.26) (0.08) Urban Primary School -0.40 0.13 (0.12) (0.09) Urban Secondary School 0.01 0.19 (0.16) (0.14) Private-Subsized Primary School 0.00 -0.01 (0.08) (0.05) Coorporation - Primary School -0.57 0.26 (0.68) (0.32) Private Primary Schoo -0.11 -0.06 (0.13) (0.08) Private-Subsized Secondary School -0.20 0.12 (0.07) (0.05) Coorporation - Secondary School -0.47 0.16 (0.28) (0.15) Private Secondary School -0.28 0.17 (0.14) (0.09) Intercept -0.36 -1.17 -0.40 (0.14) (0.30) (0.16) Unobserved Heterogeneity -0.98 -1.22 1.00 (0.09) (0.12) Notes: (a) In the case of mother's and father's education the baseline category is Primary Education. In the case of the dummies controlling for the type of management the baseline category is Public School. (b) The average score is standarized to have mean 0 and variance 1 in the population. Standard Errors are presented in parentheses. 1 Table A1. Descriptive Statistics SPS02 by Gender Variable (Dummy=1 if Apply) Mean 33.76 Age A. School Information Maximum Schooling Level = Primary Education Maximum Schooling Level = Secondary Education Maximum Schooling Level = Some Tertiary Education Maximum Schooling Level = Complete Tertiary Education A.1. Primary School Primary School in Urban Area Repeating a Grade in Primary School Was Primary School Public? Was Primary School Private-Subsidized? Was Primary School Managed by a Coorporation? Was Primary School Private? A.2. Secondary School Secondary School in Urban Area Repeating a Grade in Secondary School Was Secondary School Public? Was Secondary School Private-Subsidized? Was Secondary School Managed by a Coorporation? Was Secondary School Private? Average Grade in Secondary School Females Std. Dev. 3.76 Mean 33.71 Males Std. Dev. 3.79 0.11 0.51 0.26 0.11 0.32 0.50 0.44 0.31 0.17 0.49 0.24 0.10 0.38 0.50 0.43 0.30 0.91 0.22 0.77 0.16 0.00 0.07 0.29 0.41 0.42 0.37 0.05 0.25 0.89 0.30 0.81 0.13 0.00 0.06 0.31 0.46 0.39 0.33 0.04 0.23 0.98 0.20 0.70 0.23 0.01 0.07 0.16 0.14 0.40 0.46 0.42 0.08 0.25 0.98 0.99 0.24 0.70 0.22 0.02 0.06 -0.17 0.12 0.43 0.46 0.42 0.13 0.24 1.00 0.56 0.99 1.64 7.51 8.14 0.28 0.96 0.50 0.07 1.19 3.77 4.11 0.45 0.20 0.55 0.99 1.47 7.42 7.91 0.35 0.96 0.50 0.09 1.22 3.69 4.00 0.48 0.20 B. Family Background Mother's Employment - Asalaried Father's Employment - Asalaried Total Number of Children Mother's Education (years of schooling) Father's Education (years of schooling) Growing up under Poverty Growing up in a Broken Home C. Labor Market Variables Monthly Earnings Hours Worked per Week Hourly Wage Working During Last Month Total Work Experience since Jan. 1980 Less than 10 years of Experience Between 10 and 15 years of Experience More than 15 years of Experience 215,266 43.41 1,292 0.59 113.43 0.56 0.26 0.18 214,323 11.74 1,257 0.49 66.00 0.50 0.44 0.39 285,140 48.17 1,636 0.82 165.02 0.25 0.34 0.41 360,046 9.81 4,649 0.39 63.52 0.43 0.47 0.49 C.1 Type of Job Asalaried Employer or Self-Worker Domestic Service 0.81 0.11 0.08 0.39 0.32 0.27 0.80 0.20 0.00 0.40 0.40 0.02 Administrative and Managerial Workers Professionals Technicians and associate professionals Clerks Service workers and shop and market sales workers Skilled agricultural and fishery workers Craft and related trades workers Plant and machine operators and assemblers Elementary occupations 0.03 0.13 0.14 0.26 0.22 0.01 0.04 0.04 0.13 0.17 0.34 0.35 0.44 0.42 0.09 0.19 0.19 0.34 0.06 0.08 0.11 0.10 0.09 0.06 0.23 0.17 0.10 0.24 0.27 0.32 0.30 0.29 0.23 0.42 0.37 0.31 C.2 Type of Occupation D. Place of Residence North (I to III Regions) Central (IV to VII Regions) South (VIII to XII Regions) Santiago (Region XIII) 0.13 0.65 0.23 0.43 0.33 0.11 0.32 0.48 0.62 0.49 0.42 0.26 0.44 0.49 0.42 0.49 Number of Observations 1,765 1,801 Note: The numbers presented in this table corresponds to the sample of individuals with ages between 28 and 40 years old at the time of the interview. 2