Occupational structure and socioeconomic inequality: a comparative study between Brazil and the United States

Occupational structure and socioeconomic inequality: a comparative study between Brazil and the United States Alexandre Gori Maia Professor at Institute of Economics University of Campinas Brazil Email: gori@eco.unicamp.br Arthur Sakamoto Professors at College of Liberal Arts University of Texas at Austin. Email: asakamoto@austin.utexas.edu July 2012 No written or electronic reproduction permitted without the express permission of the author 1

Occupational structure and socioeconomic inequality: a comparative study between Brazil and the United States * Alexandre Gori Maia Arthur Sakamoto Abstract This paper investigates the role of occupational structure in socioeconomic inequality, making a comparison between Brazil and the Unites States (US). Changes in the Brazilian and American occupational structures between 1983 and 2009 are studied in order to examine if the huge structural changes witnessed in this period led to a convergence between these labor structures. Furthermore, this study considers differences in education, age, gender, race and region as sources of occupational attainment, allowing us to better understand the process that generates the higher level of segmentation and socioeconomic inequality in Brazil. The results are presented to address three general research concerns: (1) the basic conceptualization and empirical contours of occupational stratification; (2) the comparison of the trend over time in occupational stratification in Brazil and in the US; and (3) the effects of race, gender, age, educational attainment and geographic residence on occupational attainment as estimated by multinomial logit models. Overall, results highlight: (1) the higher level of socioeconomic development in the American occupational structure; (2) a tenuous trend of convergence between the Brazilian and American occupational structures; (3) the significant reduction of the impacts of education, age, gender and race on occupational attainment. In general, these results confirm the analytic usefulness of occupational structure to compare social inequality within a country over time as well as between two countries at any given point in time. Key-words Occupational structure; inequality; labor market; Brazil JEL: J21; J24; J82 * Paper presented at the 16 th World Congress of the International Labor and Employment Relations Association (ILERA), Philadelphia, US, July 2 nd 5 th, 2012. Research supported by CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior). Special thanks to Prof. Joseph Potter. 2

Introduction Occupational structure plays an important role in a diversity of studies, including those related to the level of economic development and social inequalities. Since occupation reflects the individual s position in a technical division of labor, it is taken to be a central determinant of current and future economic opportunities in modern societies (Mills, 1956; Rose and Pevallin, 2001). Occupational structure may also reflect the level of discrimination and segregation among social groups, if we believe that the occupational attainment is not just a result of individual s choice (Boskin, 1974; Brown et al., 1980; Filer, 1986; Gill, 1989; Schimdt and Strauss, 1975). Moreover, changes in the occupational structure are also related to changes in the demand for different occupational services, as a result of socioeconomic improvements and technological advances (Blau and Duncan, 1967). Since technological development affects the division of labor, it impacts on almost every aspects of the social structure (Williams, 1979). A usual feature of occupational analyses is that they are frequently based on differences of occupational attainment between social groups, such as those defined by sex, race and education. In turn, differences between countries are usually limited to the distribution of their occupational groups, without additional investigation over the level of segregation or discrimination within similar their social groups (Portes and Hoffmann, 2003; Rose and Harrisson, 2009; Szelényi, 1992). This paper analyzes the role of occupational structure as a source of socioeconomic inequalities in Brazil, making a comparison with the Unites States (US). The results are presented to address three general research concerns: (1) the basic conceptualization and empirical contours of occupational stratification; (2) the comparison of the trend over time in occupational stratification in Brazil and in the US; and (3) the effects of race, gender, age, educational attainment and geographic residence on occupational attainment as estimated by multinomial logit models. Brazil and the US are two of the main and most dynamic economies in the world, as well as they are characterized for distinct levels of social inequalities. The hypothesis under analysis is that the comparison of their occupational structures provides important elements to understand the complex level of development and inequity in these countries. Moreover, this comparison allows us identifying eventual trends of convergence or divergence in their dynamics of socioeconomic development, as well as changes in the patters of social inequalities. 1. Background: social stratification and occupational structure One of the main purposes of social stratification researches is to study the process that generates inequality and its consequences for individual behavior and social changes (Sorensen, 1996). Much of the great contribution on social stratification and differentiation theories were stimulated by the remarkable impact of industrialization on social changes and the role of class differentiation in such dynamics (Blau and Duncan, 1967). Besides other factors, occupations play an important role in defining the social structure in modern societies. Occupation defines, in large extent, opportunities of current and future income, as well as it is also related to social honor, prestige and political authority (Mills, 1956). In modern societies, political authority is largely exercised as a full-time occupation, either directly over a subordinated in a company or indirectly over others from the same social group. In this sense, some occupations provide privileged social relations for their 3

workers occupations while others are incompatible with the honor of higher prestige social groups. The study of occupational stratification presupposes an exhaustive knowledge of the occupational hierarchy, which is a major source of various aspects of social stratification in modern society (Blau and Duncan, 1967). Since occupations determine significantly the inclusion of individuals in the market and in the society, the definition of occupational classes plays a central role in a diversity of socioeconomic studies, such as historical or comparative analysis of inequality, poverty and demographic changes (Portes and Hoffmann, 2003; Sakamoto and Powers, 2003). Although such analysis does not necessarily imply the recognition of the primacy of occupations over other explanatory variables, it emphasizes the notion that occupation remains a significant and powerful determinant in many aspects of social life. Systematically, the definition of occupational classes have been based on occupational characteristics such as average wages, control over production, levels of authority, job autonomy, employment relationship, intellectual assets and skills (Wright 1985; Goldthorpe, 2000, Rose and Pevallin, 2001). Since typologies are usually based on different theoretical assumptions and present different purposes, they are not directly comparable and no assertion about a relative superiority can be easily overemphasized. Generally, any attempt to reproduce complex social relations in a discrete number of occupational categories will incur in an important loss of information. The efficiency of the occupational structure will depend, above all, on the research purpose. 2. Material and methods 2.1. Data source In order to compare the Brazilian and American occupational dynamics between 1983 and 2009, this paper uses microdata from PNAD (Pesquisa Nacional por Amostra de Domicílios - National Household Sample Survey), sponsored by IBGE (Instituto Brasileiro de Geografia e Estatística Brazilian Institute of Geography and Statistics), and from CPS (Current Population Survey), sponsored jointly by BLS (US Bureau of Labor Statistics) and US Census Bureau. PNAD is a household sample survey gradually implemented in Brazil since 1967 (IBGE, 1995). Since 1981, the whole of the Brazilian territory was covered by PNAD, with the exception of those rural areas in six Northern States, which represented less than 3% of Brazilian population 1. Recently, PNAD is based in a sample with more than one hundred thousand households (more than four hundred thousand of persons) and provides a huge and reliable source of information for socioeconomic analysis in Brazil. Inquiries on PNAD usually refer to activities conducted during the last week of September 2. CPS is a household sample survey applied monthly using a scientifically selected sample of some fifty thousand households in US (BLS, 2000). Besides providing monthly information of the labor force status in the whole US, CPS also uses supplemental inquiries to collect additional and specific information of a variety of studies (BLS, 2000). These 1 The rural areas excluded from PNAD belong to the states of Rondônia, Acre, Amazonas, Roraima, Pará and Amapá. Since 2004, these areas have been considered in the sample, but, for historical comparison, they were not considered in this study. 2 Exception for 1981, when the second week of September was adopted, and 1982, when a period between the last week on September and the second week on December was adopted 4

supplementary inquiries have the advantage of larger sample size and specific purpose design. In order to analyze specific information about socio-occupational characteristics of the American working population, this paper uses data of the Annual Social and Economic Supplement (ASEC). Information from ASEC refers to activities conducted during the week which includes the 12th day of March. Important methodological changes have been implemented in PNAD and CPS, such as those related to the classification of occupations. Between 1970 and 2001, PNAD used a common methodology to classify occupations. Since 2002, a new methodology has been used: CBO Brazilian Occupational Classification (Dedecca and Rosandiski, 2003). In turn, CPS introduced the census occupational classification on January 1971, which remained until January 1983, when occupational and industrial information began to be coded using the 1980 classification systems. On January 2003, the 2002 Census Bureau Occupational and Industrial Classification Systems were introduced in CPS (Bowler et al., 2010). Analyses in this paper considered both changes implemented in PNAD in 1983 and in CPS in 2003. In both surveys, employed was considered those who were 16 years or older and, during the reference week (a) did any work at all (for at least 1 hour) as a paid employee; worked in his own businesses, profession, or in his own farm; or worked 15 hours or more as an unpaid worker in an enterprise operated by a family member or (b) was not working, but who had a job or business from which he was temporarily absent. Persons whose job was Armed Forces were not considered in none of the surveys. 2.2. Occupational structure Although occupational classifications used in Brazilian and American household surveys are not directly comparable, special efforts were made to create common groups of analysis according. As a result, occupational codes of PNAD and CPS were aggregated in eight main occupational groups (Table 1). Some approaches were adopted to allow comparison between Brazilian and American groups. For instance, until 2001 PNAD did not have specific information for managerial position and, because of this, most Brazilian employers were considered as managerial occupations. This analysis assumes that Brazilian managers perform identical functions in the social division of labor, managing similar amounts of authority at the workplace or sharing comparable levels of prestige in relation to other occupational groups (Szelényi, 1992). Although these eight occupational classes are not social groups in the conventional sense of the term, the assumption is that their members share relatively similar life chances, social experiences and social network (Blau and Duncan, 1967). In addition, the hierarchy of occupational classes does not represent a scalar measure of social differences, but, above all, an approximation of social standards defined by the position of workers in the labor market. In other words, although some groups are subordinated to others, such relationship cannot be quantified. 5

Table 1 Occupational groups Code Occupational Group Description 1 Managers Executive, administrative and managerial occupations (excluding farmers) 2 Professionals Professional specialty, technicians and related support occupations 3 Administrative support Administrative support (including clerks) and management related occupations 4 Sales Sales occupations 5 Services Protective and service occupations, excluding private households 6 Blue collars Operators, manufacturers, laborers and precision production, craft and repair occupations 7 Private household Private household cleaners, servants, launderers and related occupations 8 Farming Farming, forestry and fishing occupations. 2.3. Multinomial logit model Besides descriptive analyses, results also considered the multiple relationships between occupational attainment and a set of personal characteristics, such as the level of education, age, gender, race and region. Two multinomial logit models were estimated for each country: one for 1983 and other for 2009. For each model, dependent variables where associated to the odds of belonging to one of the 8 occupation groups, numbered form 1 to 8 according to the codes presented in Table 1. Because the 6 th group (blue collars workers) was considered as the reference in the multinomial logit, such odds represent how larger is the probability to belong to a specific occupational group in comparison to belong to the blue collar group. In other words, the models were given by: ln( P / P x ' e (1) hi ) 6 i i h Where h=1,2,3,4,5,7,8 is the h-th occupational group, i=1..n is the i-th person, P hi is the probability for the i-th person belonging to the h-th occupational group, x i is the row vector of socioeconomic characteristics X for the i-th person, h is the vector of coefficients for the h-th occupational group and e is the column vector of unpredicted errors. The coefficients in β represent the variation in the logarithm of the odds given a unit variation in X. In order to obtain a direct relationship between a unit variation of X and the respective variation in the odds ratio, we must calculate the antilog of β, i.e., to compute e β. Explanatory factors consider: i) Education: three dummy variables in order to discriminate four levels of educational attainment College (1 for those with university degree or attending to a college or university), High Diploma (1 for those with secondary degree), High no Diploma (1 for those attending to a high school); elementary school as reference of analysis; ii) Age: represented by the decades of life (Decade) and its quadratic term (Decade 2 ); 6

iii) Female: a dummy variable valuing 1 for women, 0 for man; iv) White a dummy variable valuing 1 for white, 0 otherwise (Black, Brown, Asian, Indian, among others); v) Metropolitan: a dummy variable containing 1 for those who live in metropolitan areas; Besides predicting patterns of employment, the results of the multinomial logit model can also be interpreted as a measure of gender and race segregation (Schmidt and Strauss, 1975). Significant estimates for gender and race coefficients suggest differential access to specific occupational groups, independent of the productivity characteristics, such as age and educational attainment. However, these results must be interpreted with caution, because differences may also reflect gender and race preferences for different groups of occupations. 3. Results 3.1. Differences between occupational structures The two biggest populations and economies of America, Brazil and the US, highlight huge differences in their stage of socioeconomic development and, consequently, in their occupational structures. Some of the most expressive differences are the expressive participation of low socioeconomic status occupations in Brazil, such as agricultural and private household workers (Figures 1 and 2). Although agricultural automation reduced expressively the participation of agricultural workers in the Brazilian occupational structure between 1983 and 2009 (reduction of 13 percentage points between 1983 and 2009), they still represented 12 percent of the Brazilian employed population in 2009 (Figure 1). In Brazil, the most frequent occupation in this group is the unskilled worker in farms (38 percent of the total in this group), usually seasonal workers or unpaid workers contributing to a low productive or family agricultural production. In the US, farming workers were just 2 million of persons in 2009 and they represented just 1 percent of the occupational structure. The number of agricultural workers decreased substantially in this country between the 40s and the 80s, as a result of sharp technological growth and rising nonfarm labor returns (Barkley, 1990). Domestic workers are also usually related to informality and other precarious conditions of work. The unskilled maids comprise the majority in this group, which represented 7 million of persons in Brazil, or 8 percent of the occupation structure in 2009. In the US, domestic workers have an inexpressive share in the occupational structure, representing less than 1 million of workers in 2009. The next groups in the occupational hierarchy are the blue collars and the service workers. The blue collars, workers in manual jobs and highly routinized activities, are representative in both countries and two of the most expressive occupations in this group are the truck drivers and construction laborers. This group also presents important peculiarities in each country. For instance, in the US there are a higher participation of occupations related to industrial activities, such as laborers and assemblers. In Brazil, they are specially related to unskilled and informal activities, such as hod carriers and other manual laborers. Service occupations represent low-skilled and low-wage positions in the service sector of the economy. They differ from the blue collars more by the activities they do than the wage or social status that they have. In Brazil, they are specially related to informal activities, such as hairdressers, cooks and waiters, representing 11 percent of the occupational structure. Service workers are more expressive in the US, where they are 17 percent of the occupational 7

structure and predominantly represented by cleaners (excluding private households), cooks and nursing aides. The sharp development of markets in urban areas and big chains of wholesale and retail stores also meant a increasing participation of sales occupations in the occupational structure. They are cashiers, sales representatives, real states and a diversity of sales workers whose socioeconomic status is straightly related to the kind of customers they have. In Brazil, the expressive participation of sales workers (13 percent of the occupational structure in 2009) is determined by the large number of informal and precarious jobs in this category, such as peddlers (1.3 million of persons) and door-to-door sellers (0.4 million), representing a way out for the lack of better opportunities in the labor market for these unskilled workers. Administrative support occupations are one of the most representative groups of the white collar middle class, and they are particularly relevant in the US. Systematization reduced most of these activities to a routine and introduced a diversity of occupational categories and hierarchies in the offices. Secretaries, receptionists and auditing clerks are some representative occupations of this group, which corresponded to 16 percent of the occupational structure in the US and 10 percent in Brazil in 2009. In the upper bound of the occupational structure are the professionals and managers. Having or not direct ownership over the capital, managers run large and medium private or public firms and institutions. They control organized labor force and usually receive the highest average wages (Portes and Hoffmann, 2003). Behind them, professionals are skilled workers, usually having a university degree, who are employed by private or public institutions to assume positions of high responsibilities. In addition to administrative support occupations, managers and professionals are the dominant class in the US (50 percent in 2009), which reflect a more advanced level of development of the American occupational structure (they were just 27 percent in Brazil). Figure 1 Percent of workers according to occupational groups Brazil 1983 to 2009 Source: PNAD, microdata, IBGE. 8

Figure 2 Percent of workers according to occupational groups US 1983 to 2009 Source: CPS, March supplement, microdata, BLS The most expressive changes in the Brazilian occupational structure in this period were the reduction of agricultural workers and the rise of service workers, most of them in informal positions, such as hairdressers and cooks. In The US, changes were especially related to the reduction of blue collars and increasing of a new generation of skilled professional workers. For instance, in the US, within the professional group, the number of computer systems analysts increased by more than 1.5 million between 1982 and 2002. Since 2003, new subcategories were considered in the computer related occupation which, together, represented 3.3 million of persons in 2009. Overall, results suggest a slightly convergence of the Brazilian and American occupational structures between 1983 and 2009, especially due to the fast decrease of low status groups and increase of intermediate occupational groups in Brazil. For instance, the cumulative participation of farming, private households, blue collars and service works fell faster is Brazil than in the US. In Brazil, they shifted from 70 to 60 percent between 1983 and 2009 and, in the US, from 45 to 39 percent. On the other hand, the cumulative participation of managers and professionals increased faster in the US than in Brazil (9 and 5 percentage points, respectively). 3.2. Occupational attainment This topic analyzes the impacts of socioeconomic characteristics, determined by gender, race, age, education and region on occupational attainment in Brazil and in the US. The aim of such analyses is to identify levels of discrimination, segregation or just differences in the occupational preferences in each country. Because PNAD had no information on race in 1983, analysis for this country refers exceptionally to 1982 and 2009. In the US, analyses refer to 1983 and 2009. Table 2 (Brazil) and Table 3 (US) present maximum likelihood estimates for the multinomial logit coefficients β, as given by equation (1). Due to the high number of sampling observations, most estimates were significant at 5% significance level. Since blue collars were used as reference, estimates reflect the impact of each personal characteristic on the chance to be in a specific occupational group when compared to the chance to be in the blue collar 9

2009 Age Educ. 1982 Age Educ. group. In order to enrich analysis, the distribution of workers in each occupational group according to their socioeconomic characteristics is presented on Appendix A. Table 2 Maximum likelihood estimates for multinomial logit model Brazil 1982 and 2009 Variable Man. (P 1 /P 6 ) Prof. (P 2 /P 6 ) Adm. (P 3 /P 6 ) Sales. (P 4 /P 6 ) Serv. (P 5 /P 6 ) Private (P 7 /P 6 ) Farm. (P 8 /P 6 ) Intercept -7.15-5.08-1.77-1.13-2.14-1.21 1.82 College 4.60 5.59 4.67 2.04 2.36-1.33-0.13 + High Diploma 2.94 3.33 3.45 1.51 0.76-1.85-1.25 High No Diploma 1.80 2.10 2.41 1.20 0.52-0.80-1.66 Decade 1.77 0.76-0.74-0.51-0.24-1.55-0.93 Decade 2-0.16-0.07 0.08 0.08 0.05 0.18 0.14 Female -0.22 1.53 1.16 0.74 1.56 4.03 0.09 White 0.84 0.32 0.42 0.29-0.03 + -0.40-0.21 Metropolitan area 0.13-0.08 0.38-0.04 0.16 0.24-3.53 Intercept -5.92-4.65-1.54-0.23-1.92-3.54 1.11 College 3.48 5.14 3.82 1.65 0.84-1.37-1.02 High Diploma 1.51 2.33 2.08 0.87 0.45-0.89-1.38 High No Diploma 0.79 1.47 1.37 0.54 0.38-0.27-0.93 Decade 0.99 0.13-1.04-0.91 0.11 0.13-0.78 Decade 2-0.08 0.01 + 0.11 0.11-0.01-0.02 0.10 Female 0.66 1.47 1.56 1.40 1.26 4.05 0.11 White 0.54 0.11 0.16 0.12-0.17-0.37-0.19 Metropolitan area 0.05 0.18 0.44 0.22 0.31 0.22-2.76 + Not significant at 5% significance level Source: PNAD, microdata, IBGE First of all, the low level of education of the Brazilian working population must be highlighted. Despite having tripled the participation of workers with a secondary diploma or more since 1982, they represented just 46 percent of the Brazilian working population in 2009. In the US they were 90 percent. Within occupational groups with lower socio-occupational status, such as private household and farming workers, they were mostly inexpressive in Brazil, whereas they represented more than 50 percent in any of the American occupational groups in 2009. In both countries, the level of education shows the most expressive impact on the occupational classification. The preponderance of a consistent hierarchy of values in the multinomial logit estimates indicates that more education makes the person more likely to be in a higher occupational group as opposed to a lower occupational group, holding other 10

2009 Age Educ. 1983 Age Educ. personal characteristic constant. For instance, professionals, administrative support and managers are, respectively, the occupational groups where more educated workers are more likely to be in comparison to farming, private household and blue collar groups, where they are less likely to be. Table 3 Maximum likelihood estimates for multinomial logit model US 1983 and 2009 Variable Man. (P 1 /P 6 ) Prof. (P 2 /P 6 ) Adm. (P 3 /P 6 ) Sales. (P 4 /P 6 ) Serv. (P 5 /P 6 ) Private (P 7 /P 6 ) Farm. (P 8 /P 6 ) Intercept -7.59-6.63-4.19-1.58 1.65 1.57 0.31 + College 3.36 5.14 3.65 2.36 0.55-1.23-0.11 + High Diploma 1.47 2.01 2.40 1.11-0.09 + -1.60-0.57 High No Diploma 0.55 0.91 1.09 0.59-0.05 + -0.59-0.40 Decade 1.20 0.47-0.54-1.24-1.50-3.31-1.21 Decade 2-0.09-0.02 0.08 0.16 0.18 0.38 0.16 Female 1.02 1.82 2.79 1.57 1.75 4.59-0.22 White 0.60 0.32 0.26 0.72-0.44-0.72 0.46 Metropolitan area 0.42 0.29 0.57 0.34 0.11 0.07 + -1.70 Intercept -6.35-4.34-2.80-0.51 1.92-5.27-0.34 + College 3.46 4.50 3.24 2.13 0.27-0.64-0.92 High Diploma 1.67 1.63 1.98 1.04-0.33-1.25-1.29 High No Diploma 0.80 0.90 1.17 0.95-0.10 + -0.42-0.91 Decade 0.75-0.11-0.71-1.30-1.22-1.00-1.03 Decade 2-0.06 0.02 0.08 0.14 0.12 0.11 0.14 Female 1.60 2.25 2.77 1.92 2.18 5.26 0.49 White 0.40 0.05 + 0.04 + 0.22-0.32 0.66 0.91 Metropolitan area 0.55 0.44 0.51 0.42 0.25 0.84-1.36 + Not significant at 5% significance level Source: March supplement, microdata, BLS In the 80s, estimates pointed to a higher impact of the level of education on the occupational attainment, especially in Brazil. Such results suggest that that education used to make more difference in the occupational attainment in the 80s than in the years 2000. Since more people are attaining high school or college diploma, marginal returns of education to occupational attainment tends to decrease. Moreover, reduction of marginal returns was higher in Brazil, where there was a faster relative improvement of the educational level. The effect of age, a proxy for the working experience, on occupational attainment is not so intuitive. Most groups show a quadratic relation between age and occupational attainment, which is expressed by significant estimates for both variables Decade and 11

Decade 2. In both countries, the highest ranked occupational groups, managers and professionals, tend to present a positive relation with age. Managers also present an apparent invert U relation. In other words, holding constant others factors, the less young the worker, the more likely to be a manager in comparison to blue collar he is and, afterward, the older the worker, the less likely to be a manager he is. But the most interesting result is the expressive reduction of age differentials related to occupational attainment in both countries. Put differently, age used to make more difference in the occupational attainment in the 80s than in the years 2000. As witnessed for education, these results suggest a substantial reduction of marginal returns of age experience to occupational attainment. The share of women in the Brazilian labor force is significantly lower than in the US (6 percentage points), despite the substantial increase of 10 percentage points between 1982 and 2009. In the US, women participation increased sharply since the end of World War II (Shank, 1988), in such a way that, in 1983, women accounted for 44 percent of the labor force in this country (they were just 32 percent in Brazil). Women participation increased in most occupational groups, especially in those best classified. For instance, in 2009 women were more likely to be in the manager and intermediate groups (professional, administrative support and sales) than in the 80s. Nevertheless, substantial gap between male and female occupational attainment still persists in both countries. Holding other factors constant, women are more likely to be, above all, in the private household group. In second place, women are more likely to be in intermediary group, such as professional, administrative support, sales and services groups. Women participation is also expressively lower in the American blue collar group (14 percent in 2009). In Brazil, despite the fact that the share of the white population in the labor force is lower than that of the US, the racial gap between occupational classifications is higher than in the US. In this country, holding others factors constant, white workers are specially associated with the manager, professional, administrative support and sales groups, whereas non-white are specially associated with the private household group. In the US, white workers are more likely to be in the manager, private household and farming groups whereas non-white are more likely to be in the service group. Racial gap reduced substantially in both countries, especially in the US. Regional characteristics are also responsible for huge socioeconomic disparities and determine different patterns of spatial distribution of workers. Differences between metropolitan and non metropolitan areas are more expressive in Brazil, which could be attributed to the extreme level of underdevelopment in the rural areas and the high level of regional inequality in this country. Final considerations Occupational structure in Brazil is more polarized in less qualified occupational groups. In this country, the high participation of farming, private household workers and other types of precarious occupations reflects, among other things, an economy highly dependent on commodity production, low level of productivity and the low level of qualification of its labor force. On the other hand, upper level occupational groups (managers, professionals and administrative support) prevail in the American occupational structure and reflect an economy specialized in producing continuous flows of innovation. 12

During the 80s and the years 2000, changes in the Brazilian occupational structure were specially related to an expressive reduction in agricultural workers and the rise of service workers, most of them in informal positions. In the US, changes were especially related to the reduction of blue collars and the increase of a new generation of skilled professional workers. Overall, changes led to a tenuous convergence between the Brazilian and American occupational structures, especially due to the faster decrease of low status groups and the increase of intermediate occupational groups in Brazil. The effects of age and education on the occupational attainment reduced substantially between 1983 and 2009. These effects reduced faster in Brazil, in such a way that, in 2009, there were no substantial differentials in comparison with the US. Despite relative improvement in recent decades, educational attainment in Brazil remains substantially lower than in the US, even in lower status occupational groups. The participation of female workers is higher in the US especially in white-collar and other higher status occupational categories. Racial segregation by occupation persists as a major source of social inequality in Brazil. The role of metropolitan residence in Brazil is also larger due to the existence of poorly developed areas and consequently high rates of regional migration. Brazil is widely known for its high levels of poverty and inequality, which is strictly related to its historical process of socioeconomic development. Since colonization, and with no significant structural changes until nowadays, this country has witnessed a huge accumulation of wealth by a restricted number of persons together with high levels of social exclusion. Socioeconomic development experienced in last decades seemed to have attenuated in some extent the extreme level of social inequality in this country, especially those related to the effects of education and work experience on occupational attainment. Nevertheless, the occupational structure in Brazil reflects its relative level of economic underdevelopment in relation to its American counterpart, in addition to higher levels of racial segregation and spatial concentration of occupations, especially those related to the differences between metropolitan and non-metropolitan.areas References BARLEY, A. P. Determinants of the migration of labor out of agriculture in the United States, 1940-85. American Journal of Agricultural Economics, v. 72, n. 3, pp. 567-573, 1990. BLAU, P. M.; DUNCAN, O. D. The American occupational structure. The Free Press, 1967. BLS. Design and methodology: Current Population Survey. Technical Paper 66. U.S. Department of Labor: Bureau of Labor Statistics, U.S. Department of Commerce Economics and Administration: U.S. Census Bureau, 2000. BOSKIN, M. A conditional logit model of occupational choice. The Journal of Political Economy, v. 82, n. 2, pp. 389-398, 1974 BOWLER, M.; ILG, R. E.; MILLER, S.; ROBINSON, E.; POLIVKA, A. Revisions to the Current Population Survey effective in January 2003. Available at: <http:// www.bls.gov/cps/rvcps03.pdf>. Accessed at 28 December 2010. 13

BROWN, R. S.; MOON, M.; ZOLOTH, B. S. Occupational attainment and segregation by sex. Industrial and Labor Relations Review, v. 33, n. 4, 1980, pp. 506-517. DEDECCA, C. S.; ROSANDISKI, E. Sensos e dissensos: as inovações metodológicas do censo demográfico 2000. Revista da ABET, v. 3, n. 2, 2003. FILER, R. K. The Role of Personality and Tastes in Determining Occupational Structure. Industrial and Labor Relations Review, v. 39, n. 3, 1986, pp. 412-24. GILL, A. M. The role of discrimination in determining occupational structure. Industrial and Labor Relations Review, v. 42, n. 4, 1989, pp. 610-623. GOLDTHORPE, J. H. Social class and the differentiation of employment contracts. In: Goldthope, J. H. On sociology: numbers, narratives, and the integration of research and theory. Oxford: OUP, 2000. GRUSKY, D. B.; SORENSEN, A. B. Can class analysis be salvaged? The American Journal of Sociology, v. 103, n. 5, pp. 1187-1234, 1998. IBGE. Manual de entrevista da pesquisa básica da PNAD de 1995. Rio de Janeiro: Brazilian Insitute of Geography and Statistics, 1995. MILLS, W. White collar: the American middle classes. New York: Oxford Univ, 1956. PORTES, A.; HOFFMANN, K. Latin American class structures: their composition and change during the neoliberal era. Latin American Research Review, v. 38, n. 1, pp. 41-82, 2003. ROSE, D.; PEVALIN, D. J. The national statistics socio-economic classification: unifying official and sociological approaches to the conceptualization and measurement of social class. Colchester: University of Essex, ISER Working Papers. n. 4, 2001. ROSE, D.; HARRISON, E. (ed.) Social Class in Europe: An introduction to the European Socio-economic Classification. Routledge, 2009, 324 p. SAKAMOTO, A.; POWERS, D. A. Demography of social stratification. In: DELAMATER, J. (ed.) Handbook of the social psychology. New York: Kluwer Academic/Plenum Publisher, 2003. SCHIMIDT, P.; STRAUSS, R. The prediction of occupation using multiple logit models. International Economic Review, v. 16, n. 2, pp. 471-486, 1975. SHANK, S. E. Women and the labor market: the link grows stronger. Monthly Labor Review, v. 111, n.3, 1988. 14

SORENSEN, A. B. The structural basis of social inequality. The American Journal of Sociology, v. 101, n. 5, pp. 1333-1365, 1996. SZELÉNYI, S. Economic subsystems and the occupational structure: a comparison of Hungary and the United States. Sociological Forum, v. 7, n. 4 pp. 563-586, 1992 WILLIAMS, G. The Changing U.S. Labor Force and Occupational Differentiation by Sex. Demography, v. 16, n. 1 pp. 73-87, 1979. WRIGHT, E. Classes. London: Verso. 1985. 15

US Brazil Appendix A Percentage distribution and average values for the occupational structure according to personal characteristics Brazil and the US, 1982/1983 and 2009 Occupational Structure Middle or less High No High Diploma Diploma College Age (Avg) Female White Metropolitan Managers Professionals Administrative support Sales Services Blue collars Private Household Farming Total Managers Professionals Administrative support Sales Services Blue collars Private Household Farming Total 1982 46.5 5.2 20.6 27.7 38.6 17.0 80.5 51.0 2009 18.9 4.4 32.1 44.6 40.8 36.7 69.8 37.1 1982 28.0 6.1 22.4 43.5 33.7 55.5 73.9 46.3 2009 5.2 3.1 22.8 68.9 37.4 57.6 64.4 40.6 1982 31.1 15.5 31.8 21.6 29.1 47.4 73.4 54.4 2009 9.4 8.2 40.4 41.9 32.9 58.4 60.1 44.5 1982 78.2 8.7 9.7 3.3 34.4 34.5 64.3 40.6 2009 33.6 11.4 39.6 15.5 35.4 52.1 53.4 37.4 1982 85.1 4.6 5.3 5.0 35.0 53.9 56.1 45.2 2009 46.7 10.8 33.8 8.7 37.6 47.9 44.7 39.5 1982 93.7 3.2 2.5 0.5 33.5 19.8 55.5 40.8 2009 59.9 9.7 25.9 4.5 37.7 20.2 46.1 31.9 1982 97.1 2.2 0.5 0.1 30.3 93.1 44.7 44.0 2009 72.0 9.9 16.3 1.8 38.6 93.1 38.3 35.1 1982 98.5 0.6 0.6 0.3 36.1 20.8 49.4 1.9 2009 84.6 5.5 8.0 1.8 40.9 20.6 40.2 2.7 1982 80.8 4.4 7.6 7.3 34.1 32.5 58.2 32.9 2009 45.6 8.3 27.0 19.1 37.5 42.2 50.4 32.7 1982 1.8 4.2 27.8 66.1 41.9 31.3 95.0 71.4 2009 0.5 1.8 19.1 78.6 45.4 41.6 93.3 75.3 1982 0.4 1.4 11.7 86.5 38.3 47.3 93.7 69.8 2009 0.2 0.9 7.8 91.1 42.5 57.2 90.7 74.2 1982 0.9 4.6 46.7 47.8 37.1 75.2 91.3 72.2 2009 0.5 3.4 29.1 67.0 41.8 70.2 87.9 73.7 1982 3.1 9.1 39.8 47.9 37.0 47.6 95.4 68.1 2009 1.5 8.3 29.6 60.6 40.1 50.3 91.1 72.1 1982 10.0 18.0 43.9 28.1 35.1 55.5 83.9 62.8 2009 6.4 14.1 35.5 44.0 38.2 57.3 83.9 69.4 1982 10.7 17.9 49.9 21.5 37.1 18.9 89.3 60.0 2009 6.8 12.1 47.0 34.1 41.6 13.8 88.9 64.5 1982 28.3 34.6 26.6 10.5 33.2 95.2 80.7 61.2 2009 14.2 19.2 30.5 36.1 40.7 96.5 94.3 78.0 1982 19.9 18.5 39.1 22.5 40.5 15.3 93.8 21.3 2009 15.5 13.3 36.9 34.2 45.8 21.4 96.9 31.2 1982 6.0 10.8 38.5 44.7 37.5 44.2 90.9 65.1 2009 3.1 6.9 28.0 62.0 41.6 47.8 89.2 70.7 Source: PNAD, microdata, IBGE; CPS, March supplement, microdata, BLS 16