Reproduction of Parental Occupations, Income and Poverty in Brazil

This article aims to analyze the parental reproduction of occupations and their effects on income according to the diverse socioeconomic conditions of Brazil. Logistic regressions and wage decomposition using RIF -Recentered Influence Function—were applied to the microdata of the 2019 National Continuous Household Sample Survey. We found that sons tend to follow the occupation of fathers and daughters that of mothers; rural dwellers from traditional families and from poorer classes with low levels of education are more likely to reproduce their parents' occupation. Parental reproduction of occupations among the poor leads to lower remuneration for individuals compared to those who choose other occupations and to the maintenance of economic poverty. Impacts are greatest in the smallest quantile of wage distribution and for the poor.


Introduction
The aim of this article is to analyze the parental reproduction of occupations and its effects on the income of workers who live with their parents. 'Intergenerational reproduction', with a reputable set of works in the literature, is wider than 'parental 1 3 reproduction', the focus of our work. Actually, 'intergenerational reproduction' also considers the linkages coming from two or more past generations to the current active generation. Using microdata from the 2019 National Continuous Household Sample Survey, this paper tests the hypothesis that occupational parental inheritance adversely impacts the income of poor workers and provides for the maintenance of the economic status of poverty in Brazil.
The eradication of poverty and its deprivations in Latin America has been a priority issue and an enormous challenge for the development of countries. The eradication of poverty has been one of the goals of sustainable development pursued in the 2030 Agenda of the United Nations. According to Cepal data (2020), the poverty rate in Latin America, in the biennium 2019/2020, was observed in approximately 30% of the total population, equivalent to the values detected in 2008. The proportion of Latin American people in poverty had declined from the late 2000s to 2014, reversing this trajectory after that year. In Brazil, poverty has a multifaceted and persistent nature, and the economic and social advances achieved through the middle of the last decade have recently slowed down in a less dynamic economic context. According to information from the IBGE (2023), in Brazil, 79.4% of young people, under 29 years of age, and 33.2% of those between 15 and 29 years old, receive up to US$ 5.5 PPC 2011, the poverty line for upper middle income economies, used for Latin American countries.
On average, a worker from OECD member countries receives $49,165.00 per year. In 2019, the average annual salary of the Brazilian worker corresponded to US$ 6,272.53 (R$ 24,603.98/ US$ 3.92). For the Brazilian Poor Workers group, annual labor income was US$ 2,043.92 and for the Brazilian non-poor group US$ 7,281.98. The economic literature suggests that the persistence of poverty may be the result of inheritance but also the legacy of certain occupations (Piketty, 2000). Professional activity can condition individuals to an economically well-defined situation in the future, influenced by the family environment (Luchiari, 1996;Sobrosa et al., 2015). International studies such as those carried out by Laband and Lentz (1983), Sjogren (2000), Constant and Zimmermann (2003), Tsukahara (2007), Shilpi and Emran (2010), Thijssen and Wolbers (2016) and Bello and Morchio (2018) have advanced the relevance of parents' occupations to the occupations of their children and the repercussions for younger individuals. In Brazil, there have already been descriptive reports on the persistence of occupations and intergenerational income, discussed in Ferreira and Veloso (2006), Antigo (2010), Gomes and Cunha (2019) and Gomes et al. (2020). The contribution of these studies is in the segmentation of the working population, validating the existence of two distinct groups, the poor and not poor.
Despite the aforementioned studies, there is less explored scientific space that associates the intergenerational transmission of occupations and their impacts on the wage and poverty conditions of coresident Brazilians and, particularly, on the first economic activities of individuals in the country. Identifying the effects of occupational legacy on income for socioeconomic groups is essential to support public policies that fight wage inequality and poverty in Brazil.
In addition to this introduction, this work is organized into four more sections. Section 2 reviews the theoretical literature and empirical studies on the topic. The third section deals with the data sources and the methods used for analysis. Section 4 describes the results and interpretation of the data. Finally, the conclusions of the work are presented.

Review of Theoretical and Empirical Literature
Literature has reported that poverty is associated with occupations performed by individuals and income from work. In general, some of the theoretical arguments for economic differences between people are related to human capital, education and experience (Mincer, 1958), economic discrimination based on gender and race (Becker, 1971) and the segmentation of the labor market arising from of the worker's occupation (Azam, 2013;Doeringer & Piore, 1970;Harrison & Sum, 1979;Vietorisz & Harrison, 1973). In this sense, the segmentation of the labor market, the main theoretical support of this investigation, focuses on the economic inequalities between individuals-being these inequalities caused by the characteristics of the job, by dimensions related to different groups of workers or by the structure of the workforce.
Despite the important economic and social advances recorded in the past decade, since 2015, there has been an increase in poverty in Latin American countries, such as Brazil. The literature has also reported that poverty is associated with imbalances in the labor market but also with latent institutions in the economy. Some of the theoretical arguments for the economic gaps between workers are related to differences in the valuation of human capital, education and experience (Mincer, 1958). These imbalances have been found to be the outcome of economic discrimination by gender and race (Becker, 1971) and by segmentation in the labor market (Doeringer E Piore, 1970;Vietorisz & Harrison, 1973;Harrison & Sum, 1979) or by region of residence.
On the other hand, it has been found that the reproductive dynamics of occupations between generations cause the maintenance or aggravation of differences in remunerations. Becker and Tomes (1979) state that the income of individuals can be highly influenced by family connections in social networks, and Almeida and Melo-Silva (2011) highlighted the role of parents in their children's professional choice processes. According to Becker (1993), children with different parents usually have different incomes because the composition and level of parental investment depend on the characteristics of the children, their family connections, the parents' genetic constitutions, and the particular family culture, which involves previous kinship groups such as grandparents and great-grandparents. The existing literature, particularly exposed in Bello and Morchio (2018), points out that a worker who begins his/her career in the same occupation as his/her father has a substantially lower probability of exhibiting occupational mobility. Thus, the influence of the father on the initial occupational choice can have lasting consequences for the children in their allocation to the labor market.
Empirical works have identified the intergenerational (im) mobility of individuals' occupations and the differences in earnings in the labor market. Internationally, and from the point of view of occupations, Harper and HAQ (1997) found evidence that the family context has a strong effect on the occupational performance of individuals in Great Britain. In Germany, Constant and Zimmermann (2003) found that the family context affects professional choices not only directly through genetic legacies, social connections, and wealth but also indirectly through education. Tsukahara (2007) investigated the effects of father's occupation and his education on the occupational choice of children in Japan and suggested that children tend to choose the same occupation as their parents, and this predisposition is more evident for male children. Shilpi and Emran (2010) highlighted that parental occupation also influences their children, especially in rural contexts. Intergenerational occupational correlations between parents and children in Nepal follow the gender line (father-son and mother-daughter), but in Vietnam, both parents appear to exert significant effects on children's occupation choice. Thijssen and Wolbers (2016) found that men born into young cohorts in the Netherlands are more likely to experience downward occupational mobility than individuals born into older cohorts. They also found that cognitive skills provide individuals with significant protection against downward mobility.
The impact of occupational inheritance on wages was also studied by Sjogren (2000) in Sweden. Sjorgren (2000) found that family background influences individuals' occupational choices, thus determining their earnings prospects, through different access to investment in human capital and inheritance of parental skills. Laband and Lentz (1983), for the United States, proved that children who followed their parents' occupations had a positive wage differential in relation to those who did not, with this difference being greater for farmers, mainly due to the proximity of the house to the place of work and, therefore, the child's accessibility to the father's occupation. On the other hand, in a recent study, Bello and Morchio (2018) concluded that occupational choice in Great Britain is aligned with parental networks in 18% of cases. Occupational persistence generated by parental networks and transmission of preferences can be harmful to children, as occupational followers tend to find jobs more quickly, but they also tend to earn lower wages.
In Brazil, the professional choice of children was researched by Luchiari (1996), who additionally identified the influence parents and grandparents in the process of professional choice among their children and grandchildren. Luchiari (1996) found that, in agricultural units, the profession of parents is equal to the choice of their children for each 1 of 4 young Brazilian observed individuals. Araújo et al. (2019) analyzed how parents' occupations influence children's preferences in the Brazilian labor market. They used a discrete choice model that incorporates characteristics of individuals and alternatives of the choice set, based on the PNAD/2014 database. Araújo et al. (2019) showed that there are considerable differences between the influence of mother and father: daughters are more likely to choose the same occupation as the mother, and sons, the fathers. However, when the less wealthy do not have the opportunity to develop their skills, they do not receive a good placement in the labor market, and this is reflected in their remuneration, reproducing the mechanisms of inequality.
In terms of the wage effect, Ferreira and Veloso (2006) confirm the evidence of strong persistence of intergenerational wages. The probability of a child of a parent in the bottom quintile of the wage distribution moving to the top quintile is only 7%, while the analogous probability of a child of a parent in the top quintile is 43%. According to Pero and Szerman (2008), a source of inequality maintenance is the intergenerational transmission of income, since children of parents with given income levels tend to have equivalent levels. Antigo (2010) showed that the education and minimum wage variables play a key role in explaining the mobility of the poorest, but the intergenerational influences on children's income can be moved by other aspects of the family background.
The family economic background is also highlighted by Moura and Possato (2012) and Oliveira et al. (2003). Young people from less privileged socioeconomic classes tend not to have the same possibility of choosing occupation, because economic and institutional restrictions can limit their possibilities, which lead them to accept the household's status as a determinant.
Gomes and Cunha (2019) measured the average probability of a young person following the occupational legacy of their parents, which was 39.37% in Brazil, distinguishing among demographic groups. Gomes et al. (2020) reinforced the study for young people from rural areas and showed that male sons inherit their fathers' occupations, while daughters inherit their mothers' occupations, and the higher the level of education of young people, the greater the chances of following different professions than those exercised by parents. Guimarães, Arraes and Costa (2020) found that children of wealthy parents tend to earn higher salaries when compared to children of parents who had lower incomes, and this transmission of socioeconomic status is more accentuated at higher income levels.

Databases
The microdata used in this work were obtained from the 2019 National Continuous Household Sample Survey (PNADC) of the Brazilian Institute of Geography and Statistics (IBGE) in Brazil. The sample corresponded to 23.9 thousand employed workers aged 16 years or over living in the same household as their parents.
To identify whether or not individuals performed the same occupations as their parents, we followed the grouping strategy from the National Classification of Economic Activity (CNAE 2.0). Individuals with the same occupation as their parents were considered if they worked/exercised in the same sector as one of the parents. We chose to use the subsectors of economic activity for not generating a bias that could underestimate the percentage of young people who are influenced by their parents. If the occupation was used faithfully, the son of a "mason" who works as a "servant" could be classified as "without parental influence", whose occupations are different, but extremely correlated and are in the same sector of civil construction. The use of sectors for the 2019 base is also justified by the fact that the children are still in the same household as their parents, not having enough time for these individuals, in fact, to consolidate a profession for their lives.
The choice of the sample intended to check the legacy in the first occupations and mainly due to the lack of information about the occupations of the parents for individuals who do not live with their parents. This is a recognized limitation of the 2019 database. Therefore, only employed parents with children who were also in a paid occupation were considered, that is, unemployed, discouraged and unpaid domestic work were excluded from the sample. Given the limitation of the database, the correction proposed by Heckman (1979) was used to minimize or reduce the selectivity bias. This implies that the model could only capture information about parents who are still in the workforce.

Occupational Legacy Probability Method
The multinomial logit model has been used to measure the probability of individuals following their parents' occupations with the correction of the sample selection bias proposed by Heckman (1979). According to Heckman (1979), sample selection bias can occur due to the self-selection of individuals and the sample selection decisions made by researchers. The estimation of the probability and wage determination models incorporated Mills' inverse ratio (IMR) as an explanatory variable, calculated from the labor market participation equations. In the model the dependent variable can assume more than 2 categories: 1 for the father's legacy, 2 legacy of the mother, 3 legacy of the father and mother, 4 did not follow the occupational legacy.
The multinomial model has been widely used in this line of research, when the data on the dependent variable present a multinomial distribution, that is, having more than two categories, which reinforces its choice in this work. The model was formally introduced in Long (1997), Long and Freese (2001) and Greene (2012). The probability of occupational legacy was estimated according to Eq. (1): where y ji corresponds to the variable of occupational legacy, and j = 1, 2, 3 and 4 (the 4 types of occupation legacy-1, the father's legacy (LP), 2, the mother's legacy (LM), 3, the legacy of both parents (LMP) and 4 when there are occupations different from those of the parents, NL, the basis of comparison. is the vector of variables related to the human capital of the child and parents. This dimension was divided into low education (the baseline/reference category) -identified for those individuals with basic and elementary education, middle education (identified for those with high school), and high education (identified for those who are undergraduate and graduate). The variables Age and Age2 are used as proxies of experience.
relates to the individual's personal characteristics variables, such as being a woman (the baseline in the model is being a man) and white (nonwhite is the base).
is a vector composed of binary variables of the occupational segment of the parents, categorized between primary occupations (occupations with better working conditions and better salaries) and secondary occupations (characterized by precarious occupations and worse salaries).
is the vector of the variables identifying the economic sectors-agriculture (base), commerce, services and industry.
corresponds to the vector of location of individuals considering Brazilian large regions: Northeast (base), North, Center West, Southeast, South and Federal District. Poor is a binary variable of poverty. Individuals in extreme poverty were classified as poor, according to the multidimensional measure of poverty defined by Kageyama and Hoffmman (2006). 1 Urban is the binary variable for the individual's area of residence. Mills is the variable used to correct the selection bias, that is, the IMR (inverse Mills ratio).
Let us also detail here the division between primary and secondary occupation. For Doeringer and Piore (1970), after an individual is allocated to a certain segment, his salary depends on the internal rules of the segment regarding promotions and remuneration standard. In this approach, workers are segmented into two types of market, the primary ("primary occupation") and the secondary: the primary market, according to Doeringer and Piore (1970), is characterized by stable jobs, relatively high wages, high productivity, technical progress, there are policies for promotions based on competence and seniority within firms, as well as on-the-job training, and these jobs are normally provided by large companies. The secondary market in turn is characterized by high labor turnover, low wages, poor working conditions, low productivity, technological stagnation and relatively high levels of unemployment. Companies in this segment require and provide little or no training and, in general, a minimum of qualification is necessary. There are no promotions for competence or seniority. The firms that provide these jobs are small competitive companies that operate in a restricted market of unstable demand and access to capital is not easy, and profits are low, which makes it difficult to train labor and acquire technology. (1)

Quantile Decomposition of Income
We first estimated the equations for quantile wages, and then quantile decomposition was performed between individuals who followed (or not) the occupational legacy. We also analyzed these data to differentiate between poor and nonpoor individuals. Then, we reestimated with the correction of the sample selection bias. Wage decomposition allows us to isolate the effects of the characteristics of individuals from those factors restricted to occupational legacy and poverty. Wages were estimated using the semilogarithmic equation developed by Mincer (1958), traditionally performed by the ordinary least squares (OLS) method for the conditional mean and by the quantile regression (QR) method, initially introduced by Koenker and Bassett (1978), which allows the entire conditional wage distribution to be estimated.
The Mincerian equation described in the quantile regression model can be formalized by: where q is the vector of unknown parameters associated with the qth quantile.
Unlike OLS and maximum likelihood, the computational implementation of quantile regression uses linear programming methods. The default conditional quantile can be specified linearly by: For the j-th regressor, the marginal effect is the coefficient for the q-th quantile.
The quantile regression parameter, q , estimates the change in a specified quantile q of the dependent variable y produced by a change of one unit in the independent variable. The quantile regression method stands out from the OLS model because it offers flexibility for data modeling with heterogeneous conditional distributions. This method also presents a richer characterization and description of the data generation process; that is, this method also shows different effects of the explanatory variables, depending on the quantile of the dependent variable.
The model to estimate the determinants of earnings uses the logarithm of the hourly wage as a dependent variable. To explain the hourly wage, variables related to education and experience were used as well as variables identifying each economic sector (agriculture, commerce, services and industry) or to each Brazilian region (South, Southeast, Northeast, North, Midwest and Federal District). We also controlled for the urban level of the individual's residence, for the quality of the segment of occupation (primary versus secondary), for the formality of the job, for gender and color. Finally, the study variable related to occupational legacy was also analyzed through a binary variable assuming 1 when the individual had the same occupation as the father's, mother's or of both of them (and 0 when s/he had a different occupation than the parents).
Although the original Oaxaca-Blinder wage decomposition methodology is widely used, several works have extended the analysis beyond the original branch of contributions. DiNardo et al. (1996) discussed a methodology to decompose differences in distribution statistics beyond the mean. According to Rios-Avila (2019), this method has become known as RIF (Recentered Influence Function) decomposition and has three advantages over other strategies in the literature: the simplicity of its implementation, the possibility of obtaining detailed contributions of individual covariates in the aggregate composition, and the possibility of expanding the analysis to any statistic for which an RIF can be defined. Firpo, et al., (2018) also described the advantageous use of RIF in this methodology. We provide additional details in Appendix. Table 1 shows the characteristics of those individuals who simultaneously live with parents and follow their parents' occupations (occupational legacy) and those who followed another occupation (no occupational legacy) for 2019. In general, workers who live with their parents or guardians are between 23 and 25 years old and have a working experience of approximately 7 years. Of those who assumed the occupational legacy, 28% had low education (for the poor, this value is approximately 59%), and 15% had high education, which shows that human capital is fundamental in the possibility of individuals being in a different occupation than their own parents. For workers who did not carry out the parental legacy, these values are lower for low education and higher for high education. Men and nonwhites experience greater parental influence on occupations, especially the poor. For the poor population, the most employing sectors are the service sector in general and the agricultural sector. There are fewer legacy occupational workers in a primary occupation (which is composed of the best jobs with better conditions and payments). This means that when the parents' occupational paths are not followed, the individuals tend to be in higher paid occupations.

Description of Data
The wages and per capita household income of occupational heirs are lower than those of individuals who were allocated to occupations other than their parents. The disadvantageous situation is even worse for the poor group. A poor occupational heir has a per capita household income 6.7 times lower than that of an individual who also followed the occupational legacy but is not poor. We also observed an inverse correlation between parents' education level and the likelihood that their children will follow the occupational legacy. In relation to the area, of those who did not follow the parental legacy, the vast majority reside in urban areas, showing that the occupational legacy is stronger in the rural area, being more intense for the Poor.
Regarding the age of employed individuals who still live with their parents, 59.52% are under 25 years old and 40.48% are aged 25 years or over. Of those under the age of 25, 36% have the same activity as their parents, while 64% perform other activities. Of individuals aged 25 or over, 72.5% have occupations other than that of their parents, and 27.5% the same occupations. A more fragmented analysis can be seen in Fig. 1. Figure 1 shows the proportion of individuals who followed their parents' occupational legacy by economic class. Of the individuals who live with their parents and follow the occupational legacy, the vast majority are between 16 and 24 years of age. For the poor class, this percentage was above 77.6%, and for the nonpoor, it was approximately 62.8%, which shows a correlation between poor individuals entering the labor market early in the same occupations as their parents.
Among the poor, more than half of the employed (51.05%) had an occupation equal to or similar to that of their parents (Table 2). In general, women are less likely to follow the occupational legacy, especially if they are nonpoor; when women follow the occupational legacy, the majority assume the occupations of their mothers. Let us clarify that statistics in the Tables are weighted.
Another piece of evidence is that the higher the level of education of the parents, the smaller the occupational legacy (Table 3). Unlike the poor, the nonpoor tend to follow occupations other than their parents'. However, when parents have a low level of education, their children tend to work in professions similar to theirs.
Another relevant fact is that when individuals live in a traditional family (i.e., with both parents), they tend to follow the legacy of their parents, especially if they have been classified as poor (Fig. 2).
In general, 82.6% of the parents have a secondary occupation, and its complementary one, 17.4%. The average hourly wage of parents who are inserted in a primary occupation is R$ 35.60, and of those who work in the secondary segment is R$ 10.70. Table 4 shows that the

Probability of Occupational Legacy for Workers in Brazil
This section aims to identify and discuss the occupational influence of parents on their children's occupations, controlling for the personal characteristics of individuals and of the labor market. Table 5 presents the marginal effects of the probability of individuals following their parents' occupational legacy, with correction of the selection bias of Heckman (1979).
In Table 5, we observe the higher the worker's level of education, the lower the probability of exercising the father's occupations. For each additional year of age, there is a 7.24% lower chance of following the father's occupational legacy and a 7.69% higher chance of looking for other occupations. Women are more likely to work in the same occupation as the mother and less likely to continue in the father's occupation compared to men and to be in a different occupation than their parents. These results confirm those highlighted in studies such as Tsukahara (2007) for Japan, Shilpi and Emran (2010) for Vietnam, or Araújo  When the father has a primary occupation, it increases the probability that the children will continue in a primary occupation. If the mother has a primary occupation, the children tend to follow the mother's occupation to the detriment of the father's occupation, possibly because these activities present better working conditions and remuneration. The existing literature, particularly exposed in Bello and Morchio (2018), points out that a worker who begins a career in the same occupation as the father is substantially less likely to exhibit occupational mobility, and the father's influence in the initial occupational choice may have lasting consequences for children's placement in the labor market.
The data show that there is a lower probability of occupational legacy when individuals live in urban areas compared to rural areas, which reinforces the hypothesis that occupational legacy is greater in rural areas. This evidence is corroborated by the work of Laband and Lentz (1983) in the United States, Shilpi and Emran (2010) in Nepal and Vietnam, and Gomes and Cunha (2019) in Brazil. Occupational mobility in a rural economy is important for poverty alleviation, as mobility from agriculture to nonagriculture is often a way out of the rural poverty trap. Compared to the northeastern region, individuals from other large Brazilian macroregions are less likely to follow the legacy of parents.
Poverty influences occupational transmission between generations. A poor young person has a 25.27% probability of having the same occupation as the father and 12.49% when the parents have the same occupation. This corroborates the hypothesis raised by the vicious cycle of poverty caused by occupational transmission, constituting an unwanted legacy. Additionally, according to the condition of poverty, the poor reduce the probability of performing occupations other than that of their parents by 27.14%. In this respect, Oliveira,et al., (2003) and Moura and Possato (2012) offer similar findings, as they argue that the less favored classes tend to follow occupations that are closer to their reality and to their family.
The Mills variable was found to be statistically significant, suggesting that its insertion is relevant for correcting the sample selectivity bias. The negative signs of the coefficient indicate that there are unobserved factors that reduce the probability of assuming the occupational legacy of the parents. For the year 2019, the individual residing in the urban area was more likely to not follow the occupational legacy of the parents, 20.88% more than a worker residing in the rural area. This result reinforces the hypothesis that the occupational legacy is greater in rural areas. This can be corroborated by the urban variable,  The IIA and Wald tests were performed. The Wald Test -"Combination between categories" was elaborated; P > Chi2 values were significant at 1% significance, which indicates that the null hypothesis is rejected and, therefore, it is not advisable to combine the categories Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.  (2019) for Brazil, where the authors found that children were more likely to be employed in agriculture when the father or mother were also employed on the farm, and vice versa. The information contained in Table 6 strengthens the evidence exposed above by the probabilities of occupational legacy at the midpoints. The poor are more likely to follow the occupational legacy of their parents. Poor men are more likely to follow the father's profession, and nonpoor men seek other professions. Women tend to follow the mother's occupation. Individuals who live in traditional families and who live in rural areas are more likely to follow the occupational legacy. Rural residents are more likely to follow the occupational legacy. This converges with recent literature on the scarring effect (Bello & Morchio, 2018).
The lower the level of education, the greater the occupational legacy, which is more intense when individuals belong to the less favored socioeconomic class. A remaining remark related to Table 6, line 'single parent' -as can be checked, there is no information for PML, because, on the raw, you check a single-parent family, so there is no PML observation (PML means a home of a father and a mother living together). The probabilities of this particular case are the average probabilities, conditioned to the fact of single parenthood.

Wage Determinants
Tables 7 and 8 show the wage determinants considering the available database. Estimations for poor individuals are in Table 7. Poor individuals who followed their parents' occupational legacy receive approximately 15.24% less than those who pursued other occupations, but the disadvantages diminish at the higher quantiles of the wage distribution. This finding follows Laband and Lentz (1983), who identified a positive wage differential for individuals who carry out the same activities as their parents in the United States, and Bello and Morchio (2018), who showed that occupational persistence tends to attribute jobs more easily for children but with lower wages. A poor individual with high qualifications receives an average of 11.5% more than an individual with low education. This difference was found to be greater for the highest quantile. There is an unfavorable wage gap between men, which is minimized as wages increase. Having a formal job, a primary occupation, working in the public sector, exercising their occupation in economic sectors other than the agricultural sector, or not residing in the northeast region leads to higher wages.
With regard to the wage determinants of the nonpoor, in Table 8, there are differences in the effects found for the poor, especially in the intensity of these determinants.
Data on occupational legacy were statistically significant for the highest quantile of wages and showed a positive relationship with wages, indicating that the nonpoor, when The values presented in the table are the antilogs of the coefficients [(eβ-1)], a method known as the "Kennedy approximation", for providing accurate effect of the coefficients. See Van Garderen and Shah (2002). To minimize the heteroscedasticity problem, common in wage determination models, White's (1980) robust error procedure was used. In the general regression, the Chow structural break test was applied. F > F(critical), with 1% significance *** p < 0.01, ** p < 0.05, * p < 0. following the occupational legacies of their parents, receive higher remuneration than those who do not. This result converges with that of Laband and Lentz (1983). However, this result is different of what was found for the class of poor people, in which the occupational legacy had a punitive characteristic for individuals, which confirms the hypothesis that the occupational legacy between generations maintains the socioeconomic condition of workers.
In general, the other control variables included in the wage equation were statistically significant, and their signs are as expected. The returns are positive for the level of education and experience. Women receive lower salaries than men, and whites receive more than nonwhites. Salary increases are more likely for formal employment, public work, those who have a primary occupation, those in industry, and those who live in urban areas and in regions other than the northeast of the country. These results can be explained by the dynamics of the market, as evidenced by the theory of the labor market (Doeringer & Piore, 1970;Vietorisz;Harrison, 1973) and empirical studies for Brazil (Cacciamali, 1978;  The values presented in the table are the antilog of the coefficients [(eβ-1)], a method known as the "Kennedy approximation", which aims to present the real effect of the coefficients. For more details see Van Garderen and Shah (2002). (3) To minimize the heteroscedasticity problem, common in wage determination models, White's (1980) robust error procedure was used. In the general regression, the Chow structural break test was applied. F > F(critical), with 1% significance *** p < 0.01, ** p < 0.05, * p < 0.  Lima, 1980;Gomes, 2016;Mantovani, 2018 andPaiva, 2019). Most of these factors lead to wage gains at the highest levels of the wage distribution.

Quantile Wage Decomposition
To study the sources of wage differences between workers who assumed the occupational legacy and those who chose different occupations and the specific contribution of occupational transmission to workers' wages and socioeconomic status, we recurred to the RIF decomposition of wage differences into quantiles presented by Rios-Avila (2019) and already used by Scicchitano (2012). RIF decomposition is an extension of the Oaxaca-Blinder decomposition that disaggregates wage differences into a portion explained by differences in workers' characteristics and an unexplained portion, which can be understood as a proxy of the occupational legacy effect on workers' wages when comparing the wages of the occupational legacy and nonlegacy groups (or a proxy of the vicious cycle of poverty when comparing the poor and nonpoor groups). Table 9 shows the wage decomposition of people associated with occupational legacy and nonlegacy. For the group of poor people, occupational legacy presented the greatest penalty (larger wage differences) for workers with the lowest quantile (q25). For the wages of individuals from the occupational legacy to be equal to those of workers from the nonoccupational legacy, their wages should increase by 49.83%. Part of this difference (14.07%) is explained by differences in the characteristics of individuals or the job position. If the individuals from the occupational legacy had the same characteristics as the nonlegacy, their wages would increase by 10.08% (at q25). This confirms the hypothesis that nonoccupational legacy individuals have better remuneration characteristics. For the higher quantiles, this explained difference is smaller, which confirms an aggravation for the poor who followed the occupational legacy and who are in the lower quantiles of the wage distribution. With regard to the unexplained part, if individuals had not followed their parents' occupation, they would have received an increase in their wages of 26.98% for the lowest quantile (q25), with a smaller impact on the highest quantiles. This shows that the occupational legacy, for the poor group, penalizes workers with lower wages.
Differences in disfavor to the occupational legacy group are also observed for the nonpoor working class. However, in contrast to the poor group, this difference is greater for the higher wage quantiles. With regard to the explained part by the base model, this also showed that for higher wage quantiles, the explained part is more significant.
The occupational legacy effect was more significant for the poor class than for the nonpoor class. If the poor worker had not followed his parents' occupational legacy, her or his salary would have increased by 3.89% when these belonged to quantile 25 of the salary distribution. For the higher quantiles, the conclusion is inverse: not following the parents' legacy would reduce individuals' wages. Again, occupational transmission between generations was found to be associated with the persistence of poverty or wealth in Brazil. Table 10 analyzes the effect of poverty on occupational heirs and nonheirs in Brazil, measured by quantile wage decomposition. First, the gap between the wage incomes of the poor and nonpoor is smaller for the higher quantiles of the wage distribution. If poor heirs had the same characteristics as nonpoor heirs, they would observe an increase from 28.89% (q25) to 36.35% (q75), and most of this difference is explained by the base model. For those who did not inherit their parents' occupations, this additional gain was between 20.19% (q25) and 38.68% (q75). The higher the wage quantiles, the greater the gains resulting from the workers' profiles (which also follows Boar and Lashkari, 2021).
The cause of wage inequalities specifically caused by the poverty of heirs and nonheirs, the unexplained part, can be considered the proxy of the vicious cycle of poverty. In a counterfactual analysis, if the poor individuals who inherited their parents' occupation were not poor, their salary should increase 100% for those with lower salaries and 39.10% for those with higher salaries (check estimates for the last row in Table 10, 'Pure . For those who did not follow their parents' occupation, these impacts were smaller, and the trend of reduction in these returns continues along the wage distribution. The effect of poverty on the wage differentials of individuals is relevant; it goes beyond production and market factors, decreases along the distribution of wages and is greater for workers who carry out the same work activities as their parents. These results suggest that poor individuals who follow the occupational legacy enter a vicious cycle of poverty, with the maintenance of a disadvantaged socioeconomic situation. For the occupational legacy of the nonpoor, there tends to exist a vicious cycle of wealth. Greater upward income mobility for the poorest people can break the vicious cycle of poverty to which these individuals are subjected. If this occurs, it is possible that the country, in the long term, will present greater income equalization (Antigo, 2010).

Conclusions, Implications and Further Work
This research studied the probabilities of children reproducing parental occupations in Brazil. We studied those workers living with at least one parent. First, we observed that individuals who live in the same household as their parents and perform the same occupation as their parents are just over 23 years old, male, nonwhite, with low education, with low education parents, are in secondary occupations, with lower salaries, and reside in rural and northeastern Brazil. These characteristics are more intense and unfavorable among the poor.
The poor are more likely to follow the generational occupational legacy, while the nonpoor pursue different occupations. Women tend to follow the occupations of mothers and men, their fathers. The legacy was more evident in rural areas and smaller in single-parent families. The level of education of individuals reduces the chance of an occupational legacy, as the higher the worker's education, the greater the chances of pursuing professions other than those pursued by their parents.
The probabilities of occupational transmission between generations differ between economic classes, unfavorably for the poor who still live with their parents. The occupational legacy for the poor can condition individuals into a "poverty trap".
The results related to the determination of salaries show that there is a salary difference between individuals who followed their parents' occupations and those who chose different jobs, the former with the lowest salaries. The impacts on wages controlled for individual characteristics are different for the analyzed wage distribution quantiles. Schooling was shown to have smaller impacts for the first quantile. Conversely, the schooling effect was greater for the highest quantiles in the distribution of wages. The occupational legacy variable proved to be more homogeneous across the distribution for the poor than for the nonpoor.
With regard to wage decomposition, there is a wage differential that is attributed to the occupational legacy, which also exists as a penalty for poor individuals and a bonus for nonpoor individuals. Most of the difference can be explained by the human capital of individuals or by the segmentation of work. When analyzing the groups of poor and nonpoor individuals, it is also possible to observe an unexplained part of the wage differential, which is considered a proxy of the vicious cycle of poverty. In a counterfactual analysis, if individuals did not belong to the class of the poor, they would see an increase in their wages.
Thus, the results confirm the hypothesis raised that individuals who tend to follow their parents' occupations earn, on average, less than those who chose different occupations from their parents. This fact is a source of the poverty trap, as economically disadvantaged individuals, generally younger, tend to follow in their parents' footsteps, and often, in parent-child cohabitation, they must help with household activities and contribute to household income. Forced to opt for work or study, they tend to interrupt their education prematurely, enter the labor market prematurely, and perform secondary occupations.
Thus, public policies are of paramount importance to promote continuing education as an opportunity for inclusion in undergraduate courses in public and private networks, in addition to technical and professional courses for the less favored economic class.
Access to the first job should be fostered by labor market policies, promoting more options, especially for the youngest, giving them the right to choose to escape the unwanted legacy and avoid the vicious cycle of poverty. Therefore, the policies for the insertion of workers from the less favored classes in universities and technical courses are important, since that way they will be able to occupy positions with better working conditions and higher salaries, in general, different from their parents.
We are aware that a work of this nature poses exciting challenges for further work. We propose four derivations. First, as soon as possible, carry out comparative work with databases from other South American economies to assess the weight of family inheritance in these cases. Second, we suggest greater detail at the level of the Brazilian regions, seeking to explore the additional weight of spatial autocorrelation in the evolution of wage inequality. Third, as currenty there is no available and comparable information regarding parents' education as well as parents' employment income (both of which could potentially explain the occupational opportunities of the child), we also recognize these dimensions as relevant control variables in our models as soon as they appear in a robust data source. Fourth, it is still an open debate whether comes first -the poverty level of a household or the trend of an occupational transmission. Although there appeared some methodological and empirical approaches on the topic, we consider this discussion a relevant for being developed in further researches.Finally, we propose, as soon as the databases allow, to include the weight of the long generational inheritance, examining the connection between the occupations and the remunerations of grandparents and great-grandparents in the earnings of current descendants.

Appendix: The Recentered Influence Function (RIF)
Suppose that there is a joint distribution function that describes all the relationships between the dependent variable Y, in this study the logarithm of the hourly wage, the exogenous characteristics X and the categorical variable L: f Y,X,L y i , x i , L i . Since there are only two groups based on L, the joint probability distribution function and the conditional Y cumulative distribution on L can be written as: where the subscript k indicates that the density is conditional on L = k with k ∈ [0,1]. To analyze the differences between Groups 0 and 1 (legacy and nonlegacy) or (poor and nonpoor) for a given distributive statistic v, the conditional cumulative distribution of Y can be used to calculate the difference (7): To identify differences in characteristics (explained effect) and differences in coefficients (wage structure effect or coefficient effect) to explain the overall difference in Δv (distributive statistics), it is necessary to create a counterfactual scenario.
The counterfactual statistic is defined as follows: The difference in the v distribution statistic can be disaggregated into two components: where Δv X reflects the difference attributed to the differences in the characteristics, while Δv S reflects the differences attributed to the relationships between Y and X, that is, the difference in the coefficients. The difficulty is the identification of counterfactual statistics v c because the combination of characteristics and results is not observed in the data. The term Δv X constitutes the difference in wage returns that the two groups receive due to their labor market characteristics (i.e., the counterfactual distribution), while the term Δv S is the effect of differences in returns on the characteristics between legacy and nonlegacy occupational and between poor and nonpoor. Counterfactual statistics v c can be misidentified, as discussed in Barsky et al. (2002). An alternative to correct this possible bias is to use a semiparametric reweighting approximation, as discussed in Barsky et al. (2002) and DiNardo et al. (1996), to identify the counterfactual distribution based on observed data.
The problem with identifying the counterfactual scenario is that it is not possible to directly observe the distribution of outcomes and characteristics that the counterfactual distribution F c Y|X implies. However, from an abstract point of view, it is possible to obtain an approximation for the counterfactual distribution by multiplying the observed distribution of characteristics ( dF 0 X (X) ) by a factor (X) so that it resembles the distribution dF 1 X (X) , as shown in (5).
Using Bayes' rule, the weighting factor (X) factor can be identified as follows: where P is the proportion of people in Group L = 1 and P(L = 1|X) is the conditional probability that someone with characteristics X is part of Group 1 (i.e., that someone is assumed to have an occupational legacy). In other words, to identify the counterfactual distribution F c Y|X , one can estimate the reweighting factor (X) using parametric (or nonparametric) methods to estimate the conditional probability P(L = 1|X) . Once these weighting factors are obtained, the equation is estimated as: Decomposition components can be defined as: The components Δv P S + Δv e S correspond to the effect of the aggregate wage structure of Oaxaca-Blinder (unexplained part), while Δv P X + Δv e X corresponds to the effect of the aggregate composition (explained part). These two components are decomposed into a pure wage structure ( Δv P S ) and compounding effect ( Δv P X ), in addition to two components that can be used to assess the overall adequacy of the model. Δv e S is the weighting error used to assess the quality of the weighting strategy and is expected to be zero in large samples. Δv e X is the specification error, used to assess the importance of deviations from linearity in the model specification or in the RIF approximation, according to Rios-Avila (2019). This is a semiparametric method, and the quantile regression framework does not need any distributive assumptions and allows the covariates to influence the entire conditional distribution. To estimate standard errors and confidence intervals, the bootstrap method is used.