Policy Research Working Paper 8999 Long-Term Study of PROSPERA on Intergenerational Occupational Mobility Iliana Yaschine Delfino Vargas Curtis Huffman Hiram Carreño Ulises Hernández Tlacaelel Mendoza Social Protection and Jobs Global Practice September 2019 Policy Research Working Paper 8999 Abstract Two decades after the inception of Mexico’s conditional of the youths, although their education, first occupation cash transfer program, PROSPERA, this study analyzes the and cognitive abilities are factors that, altogether, have a intergenerational occupational mobility and occupational greater weight and may reduce the effect of social origins attainment of a group of rural beneficiary youths between on occupational destinations. Women and migrants present ages 18 and 35 years, segmented into subgroups by sex, the highest rates of upward mobility and greater equality in ethnic background and migratory status. Furthermore, it labor opportunities, compared to men and non-migrants, evaluates if a higher intensity of PROSPERA’s treatment respectively. No differences due to ethnicity were found. increases the equality of labor opportunities for the youths. The findings on the effects of PROSPERA suggest that Half of the youths achieved upward mobility relative to higher levels of treatment intensity may generate greater their occupation of origin, but, at the same time, there probabilities of better occupations, although this effect is also was a high probability of having an occupation in a considered modest. The results are only valid for the ana- lower stratum of the occupational hierarchy, experiencing lyzed subpopulation and reflect a reduced difference in the high occupational inheritance and barriers to climbing the treatment intensity, which must not be considered as the social ladder. The variables related to social origin have a complete effect of the program’s intervention. significant correlation with the occupational destinations This paper is a product of the Social Protection and Jobs Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at ilianaya@unam.mx and the coordinators of the studies may be contacted at cavilaparra@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Long-Term Study of PROSPERA on Intergenerational Occupational Mobility1 Iliana Yaschine, Delfino Vargas and Curtis Huffman2 In collaboration with Hiram Carreño, Ulises Hernández and Tlacaelel Mendoza Key words: conditional cash transfers; intergenerational occupational mobility; occupational attainment; latent treatments; impact evaluation JEL: I38; C31; J62; J68; Z13 1 This paper is part of the research project “Studies of PROSPERA’s long-term results” which was funded by the Government of Mexico through the National Coordination of PROSPERA (CNP) as part of a technical cooperation agreement with the World Bank. The World Bank team is grateful to the Government of Mexico, including the Ministry of Social Development (now Ministry of Welfare), Ministry of Education, and Ministry of Health for the close collaboration in this research project, including the provision of the main datasets. The World Bank team would also like to thank the National Council for the Evaluation of Social Development Policy (CONEVAL), the National Institute of Public Health, development partners such as the Inter-American Development Bank, and the national and international researchers that contributed in the Advisory Evaluation Group. This group was established to ensure technical quality and methodological rigor of the “Studies of PROSPERA’s long-term results.” These studies were coordinated by Maria Concepción Steta, WB Senior Social Protection Specialist; Clemente Ávila, WB Social Protection Economist; and Mónica Orozco, WB Consultant.  The authors thank Clemente Ávila, Raymundo Campos, Rodolfo de la Torre, Polly Jones, Mónica Orozco and Concepción Steta for their comments, as well as the participants of the two seminars organized by the World Bank and PROSPERA, where the progress of the research was discussed. Furthermore, we acknowledge the work of the PROSPERA and the World Bank teams for integrating the ENCEL panel that we used. 2 University Program of Development Studies (PUED), National Autonomous University of Mexico (UNAM). I. Introduction PROSPERA was introduced in 1997 as one of the main programs of Mexico’s federal social policy.3 Throughout two decades of operation, this conditional cash transfers program (CCT) consolidated as an essential piece of the social policy of the country. As of the end of 2017, it has reached a coverage of 6.6 million households (27 million people), which represents around 22.7% of the national population, with a presence in 114,000 localities in all states of Mexico, in both rural and urban areas (Presidency of the Republic, 2018). The budget of the program for that year represented 0.36% of the GDP and 21.3% of the expenditure on poverty alleviation (Presidency of the Republic, 2018). As one of the pioneer CCT programs, PROSPERA has played a transcendental role at an international level. Its intervention model has influenced the design and implementation of various programs in all regions worldwide and, in this regard, has been key to the proliferation of CCTs, as preferred instruments in social policy.4 Whereas the achievement of the program’s objectives in the short and medium term has been thoroughly studied, it has become crucial to analyze the extent to which it has reached its ultimate purpose: to contribute to the breaking of intergenerational transmission of poverty. The evaluation of the long-term effects of the program is of vital importance, not only to know to what the extent PROSPERA has achieved its objectives and ultimate purpose, but also to provide useful evidence for other CCT programs. In this context, this study seeks to contribute to this exercise through the study of the intergenerational occupational mobility of the youth beneficiaries of the program in rural areas. The analysis on this subject is considered an approximation that will contribute to our understanding of extent to which PROSPERA has achieved its ultimate objective, which is an improvement of the occupation of youths, in comparison with their characteristics of origin, encouraging the process of breaking the intergenerational transmission of poverty. The data from the Evaluation of Rural Households Survey (ENCEL) collected in 2017 enabled us to study the intergenerational occupational mobility and the occupational attainment process experienced by the beneficiary youths of PROSPERA two decades after the intervention began and, at a more advanced moment in their lives, in which a greater proportion of them has entered the labor market and achieved a more stable labor status. Based on a quantitative analysis, which uses the ENCEL 2017 as the main source of data, the objective of this study is to provide answers to two sets of research questions: 1. What are the characteristics of the intergenerational occupational mobility in the rural youth beneficiaries? Do differences by sex, ethnic background or migratory status exist? What is the                                                              3 Throughout its history, the program has been named Program of Education, Health and Nutrition (Progresa), Oportunidades Human Development Program (Oportunidades) and currently Prospera, Social Inclusion Program (PROSPERA). The name PROSPERA will be used in this paper, except for those occasions in which it is more appropriate to use the name corresponding to a specific historic period. 4 Currently, there are 64 similar social programs in different regions of the world (World Bank, 2018c: 37). 2 effect of a higher intensity of PROSPERA’s treatment on the youth’s intergenerational occupational mobility? 2. What are the determinants of the occupational attainment of the rural youth beneficiaries? Do differences by sex, ethnic background or migratory condition exist? Addressing the first set of questions will enable us to identify how much the social stratum of the rural beneficiary youths of PROSPERA has changed relative to their providers (usually their parents), as well as the strength and pattern of the association between their occupational origins and destinations. Furthermore, we will explore if this varies by sex, ethnic background or migratory status, as well as if a higher intensity of PROSPERA’s treatment increases intergenerational mobility and reduces inequality of opportunities.5 The analysis of these questions is carried out through the construction of intergenerational occupational mobility tables, the calculation of the absolute mobility rates and the application of log-linear models to analyze relative mobility or inequality of opportunities. The analysis of the second set of questions will identify some of the factors, and the strength to which they do, that affect the occupational attainment of rural beneficiary youths of PROSPERA and that may help to have a better understanding of the results of the first set of questions. It is important to identify how much weight some ascribed factors related to their social origin (the constructs of social origin and the provider’s cognitive abilities) and non-ascribed factors that act as mediators between social origin and occupational attainment (the constructs of the youth’s cognitive abilities, education and first occupation) have on the youths’ occupational attainment. Likewise, it will allow us to know if other ascribed characteristics such as sex and ethnic background shape this process and whether migration from their locality of origin can reduce the weight of the ascribed factors. This document is structured into seven sections, beginning with this introduction. Section II contextualizes the design, impacts and limitations of PROSPERA within the macroeconomic and social development context of Mexico during the past two decades. Section III provides the analytical framework that guides our research, based on the studies of intergenerational social mobility, specifically those within the occupational area. Section IV describes the methodological design of the study, including the description of the study group, the proposed design to assess the characteristics of the program, as well as the variables and methods used for the empirical analysis. Section V presents the results related to the first set of questions, referring to the characteristics of intergenerational occupational mobility of the rural youth beneficiaries and the effects of the intensity of PROSPERA’s treatment in this context. Section VI does the same for the second set of questions on the determinants of the occupational attainment of this group of youths. Finally, we conclude with a summary of the findings and some closing remarks derived                                                              5 Equality of opportunities is a conception of social justice that enables free access to different socioeconomic positions for any individual regardless of the socioeconomic position of origin or other factors such as sex or ethnic backgrounds, since the core premise is that their own merits are sufficient for achieving a person’s rightful place in society (Turner, 1986; Breen and Jonsson, 2005; Dubet, 2014). Therefore, a policy of equality of opportunities will make it so that the degree to which individuals will achieve the objective (welfare, education, income, etc.) is independent from their circumstances, and only dependent on their effort (Roemer, 2003). 3 therefrom. II. PROSPERA in the Mexican macroeconomic and social context Since the mid-1980s, within the context of a serious economic crisis, the country has reached a turning-point in its development strategy, which translated into a modification in the balance between the role of the state and the market than had prevailed in the previous decades. The size of the state was reduced through the privatization of government-controlled companies and the driver of growth was steered toward the external market, which was associated with rapid trade and financial liberalization. Hence, it was considered that not only macroeconomic stability would be achieved, but also growth in the economy and an extension in the labor demand, which would translate into better levels of welfare for the population (Moreno-Brid and Ros, 2010; Cortés and Rubalcava, 2012). One of the outstanding features of the new development model in the social policy area was the strengthening of targeted programs as intervention instruments, as they were considered more efficient than the prior interventions (such as generalized subsidies) which aimed to overcome poverty. Since the late 1980s, they have begun to consolidate as an increasingly important piece of the national social policy, aligned with the current ideological stance, given the weakness of the development model to generate economic and social benefits for all the population, and within a welfare regime characterized by segmentation and stratification of its institutions and benefits (Valencia, Foust and Tetreault, 2012; Yaschine and Ochoa, 2016). In the mid-1990’s, Mexico faced another important economic crisis, as well as a political and social crisis. The government led by Ernesto Zedillo (1994-2000) faced, among other challenges, restoring macroeconomic stability, promoting growth, improving the welfare of the population that had been affected by the crisis and recovering the political legitimacy. On economic matters, the government of Zedillo continued with the economic policies of his predecessors. In the social area, in 1997 PROGRESA was launched as the pillar of the federal government’s targeted actions and as a central piece of the social development strategy (Presidency of the Republic, 1997). At that time, the required budgetary conditions to launch the program already existed, derived from the economic recovery and the reorganization of the public expenditure towards policies which were considered more efficient (Levy and Rodríguez, 2004; Cortés and Rubalcava, 2012). The design of the program was based on a diagnosis of the situation of poverty at the national level, as well as the analysis of the determinants of poverty and its intergenerational transmission. The intervention’s proposal was supported by the identification of low levels of human capital as being a core impediment to improving the level of welfare of families living in conditions of extreme poverty. In this regard, from the outset, the program proposed to implement actions to promote the development of human capital (education, health and nutrition) of the members of these families, mainly the younger generations, as a mechanism for the contribution to breaking the intergenerational transmission of poverty (Progresa, 1997; Levy and Rodríguez, 2004; 4 Hernández, 2008). Based on the analysis performed, the program set out two interrelated objectives: 1) improve the families’ welfare by improving their consumption capacity, and 2) develop the human capital (education, health and nutrition) of its members, mainly children and youths, as a mechanism for the improvement of their welfare in the future. The central hypothesis guiding the design of the program is associated to this second objective: the investment in education, health and nutrition of the new generations would enable them, when they became adults, to join higher status activities in the labor market, with greater productivity and remuneration. This would promote equality of opportunity, social mobility and would contribute towards achieving the ultimate purpose of the program: breaking the intergenerational continuation of the poverty cycle (PROGRESA, 1997). The program’s intervention used the delivery of cash transfers to the selected households, in- kind transfers and conditionalities to achieve its objectives. In recent years, since the establishment of the linkage component in 2016, an effort has also been made to link the beneficiaries of PROSPERA with other social programs in order to promote their labor, financial and social inclusion, as well as their productivity. Within the framework of the program, the achievement of its ultimate purpose required a favorable macroeconomic context and dynamic labor markets, as well as the provision of quality education and health services. In reality, PROSPERA has operated in an adverse economic and social context, which has imposed restrictions to its potential contribution to the interruption of the inheritance of poverty. The results that have been achieved throughout this period have not been at the desired level, not only have there been recurrent crises, but the dynamism of economic growth, in recent years, has been lackluster, with an annual average rate of GDP growth, in real terms, of 2.6% (Moreno-Brid, 2016: 43) and a GDP per capita growth rate of 1.09% between 1987 and 2015 (World Bank, n/d). Low economic growth, increased openness of trade and labor flexibilization have led to precarious labor markets and to the depletion of formal job creation, in favor of informal positions (Martínez and Rosales, 2018; Ochoa, 2018). This has resulted in an increase in the vulnerability of the workers in occupations with low productivity (Ochoa, 2018). These unfavorable conditions are aggravated for youths, who face a labor market with low levels of labor demand, jobs that are insecure, heterogeneous and with bad working conditions, higher rates of unemployment, lower earnings and worse labor conditions (Mora and Oliveira, 2009; Mora and Oliveira, 2012; Navarrete, 2012; Navarrete, 2018).6 On top of the weaknesses of the labor demand and the lack of linkage between the education and productive sectors, the youths’ own characteristics make a favorable insertion in the labor market more difficult, among which we find weaknesses in their schooling, work experience, as well as technical and soft skills (Ortega and Cárdenas, 2017).                                                              6 In 2017, for youths between the ages of 15 and 29, who were employed, 59.5% had informal jobs, whether for being in the informal market or for working in the formal sector without social security benefits (Ortega and Cárdenas, 2017). 5 The agricultural sector clearly illustrates the consequences of the low dynamism of the Mexican economy. Since the increase in trade openness, the employment in rural areas considerably dropped (Samaniego, 2018) which is explained by the increase in imported agricultural products (Luiselli, 2017). Furthermore, the land ownership reforms in the 1990s further deepened the sector’s deterioration; 76% of the agricultural and forestry production units are smallholdings, which provide little or no profitability and that, by and large, are dedicated to self-consumption (Luiselli, 2017: 237). This is the type of unit which experienced the most growth in the agricultural sector and which generates most employment in the sector (56.8%) (Robles, 2016). However, it is important to mention that the share of wages and cash transfers (public and private) has increased in the total income of rural households (Yúnez, 2010: 12). The national welfare system, due to its limited and unequal provision of goods and services to the different strata of the population, has hardly been able to offset the adverse effects of this static economic environment on the quality of life of the population, although it has mitigated them.7 In this context, there have been persistent levels of inequality and poverty, as well as a greater rigidity in the social mobility of the country. The Gini coefficient of the income per capita has slightly increased between 1984 and 2014, from 0.489 to 0.508 (Cortés, Ochoa, Vargas and Yaschine, 2018: 273). The prevalence of poverty has not changed significantly. Over half of Mexico’s population lived below the asset poverty line in 1992 and in 2016 (CONEVAL, n/d),8 while the incidence of multidimensional poverty in 2008 and 2016 was around 44% of the national population (CONEVAL, 2018) ).9 In both cases, an increase in the number of people living in poverty according to each definition is observed. In the case of social mobility, according to Cortés and Escobar (2007), Solís (2007), Solís, Cortés and Escobar (2007), Zenteno and Solís (2007) and Solís, (2016), the current economic model has reduced opportunities for occupational attainment in all social strata compared to the previous period associated with industrialization through an Import Substitution Economic Model. Furthermore, according to these aforementioned authors, the distribution of these opportunities among the population has been more unequal than in the past, because the reduction has been greater among the individuals from the lowest socioeconomic strata. This demonstrates an increase in inequality of opportunities and the rigidity of the national social mobility system. The analysis of the individual occupational stratification process of Mexicans has shown that, although education is the factor with the highest effect on occupational attainment, variables related to the social origin also have an important influence (Puga and Solís, 2010; Solís, 2007). Furthermore, it is highlighted that in Mexico, having a rural origin is an additional factor of disadvantage for educational and labor attainment (Puga and Solís, 2010). In this context, school                                                              7 Cortés (2018) shows that the set of public cash transfers, including those of PROSPERA, lessened the effect of the poorly performing economy and rising food prices on the incidence of poverty. 8 Assets poverty refers to a situation in which it is not possible, with total current income, to meet minimum food, health care, education, clothing, housing and transportation requirements, even if all of a household’s total current income was used exclusively for the acquisition of these goods and services (CTMP, 2002). 9 As of 2008 the official measurement of poverty in Mexico has been guided by the multidimensional methodology published by CONEVAL (CONEVAL,2018). 6 quality is an important factor since it is directly related to the capacity of the educational system to impact the life opportunities of the population (Blanco, 2011). It has been documented that inequalities in learning outcomes are associated with the type of school attended. For example, according to Saravi (2009), at secondary level of education, private schools have performed better and, among the public schools, those of a general modality (secundarias generales) have better results. Although the coexistence and interrelation of ascribed factors and education in the occupational attainment process concur with evidence from international findings, in the case of Mexico, research results have suggested that the effect of schooling on occupational attainment may be declining in recent decades, while the effects of the other ascribed factors are increasing (Puga and Solís, 2010; Solís, 2007). This further demonstrates the increase in the inequality of opportunities and the rigidity of national social mobility. The most recent studies of intergenerational mobility at a national level have confirmed the rigidity of the Mexican social stratification. Mexican society is characterized by high rates of social mobility in the middle sectors, but high levels of persistence for both the upper and lower sectors, which contributes to the reproduction of wealth and poverty. Individuals face difficulties moving to different social stratums than the one they were born into. It is extremely hard to experience long-distance changes (for example, having a father who is an agricultural worker and becoming a non-manual professional, which represent extremes of the occupational hierarchy) or to cross the existing barrier between the rural and urban sectors (Torche, 2010; CEEY, 2013; Solís, 2016).10 Furthermore, inequalities between men and women have been documented. Notwithstanding females experience a greater upward mobility, in comparison to males, those born into the lowest strata remain more frequently in lower positions, while the females with origin in the higher strata tend more to experience downward mobility (CEEY, 2013). This rigidity not only characterizes the intergenerational mobility of individuals, but also that of the territories. Cortés and Vargas (2017), based on analysis of the evolution of the municipalities in Mexico between 1990 and 2015, identified a tendency of immobility in municipal marginalization levels. Valdés and Vargas (2018) confirmed this pattern of immobility analyzing the evolution of the municipal social backwardness between 2000 and 2015. Even within this adverse context, PROSPERA has been able to contribute to well-being improvement of its beneficiaries. It has been documented that the program has had positive impacts on different human capital indicators.11 In the education area, of specific relevance for this study, PROSPERA has reduced school dropout rates for youths from rural localities, and has improved early enrollment to school, transition from primary to secondary school, and consecutive grade progression, as well as the number of years of schooling received (Parker, 2005; Parker, Behrman and Todd, 2005; Todd et. al., 2005; Parker and Behrman, 2008). Other                                                              10 In comparative terms, it has lesser social fluidity than countries such as Sweden, United States, Brazil and Chile (Torche, 2010). 11 The number of studies documenting impacts on human capital is very large. The following are summaries of such impacts in different historic moments of the program, IFRI (2000), Cruz, de la Torre and Velázquez (2006), Campos (2010) and Parker and Todd (2017). 7 recent studies have demonstrated an increase of either 1.4 years (Parker and Vogl, 2018) or 3 years (Kugler and Rojas, 2018) of average schooling, resulting from the program’s intervention in rural areas. However, the lack of quality of the education received by beneficiaries (Agudo, 2008; González de la Rocha, 2008; Mancera, Serna and Priede, 2008) and limitations regarding the learning level achieved (Mancera, Serna and Priede, 2008; Mancera, Priede and Serna, 2012) have also been highlighted. Hence, it is valid to ask if the magnitude of these impacts and the quality of human capital generated are sufficient enough for the program to achieve its ultimate purpose. Studies that have estimated the program’s impact, after 10 years of exposure, on the new generations of youth entering the labor market in rural areas (González de la Rocha 2008; Rodríguez-Oreggia and Freije, 2008; Ibarrarán and Villa, 2010; Rodríguez-Oreggia, 2010) and their intergenerational occupational mobility (Rodríguez-Oreggia and Freije, 2008; Rodríguez- Oreggia, 2010; Yaschine, 2015) have presented mixed results. Due to the length of the program’s intervention period, such studies analyzed youths who had not reached an age to have labor results that could be considered stable and representative of their adult lives (a range from 18-24 years of age). The analyses both affirmed that the low economic performance of the country and the context of labor insertion into which the youths entered was unfavorable for achieving better outcomes, due to the limited and uncertain employment options and the agricultural crisis. Regarding the analysis on labor insertion, Rodríguez-Oreggia and Freije (2008) indicated that the labor insertion of the youth beneficiaries of the program is precarious and that, while the program had a positive impact on the income of males with primary and secondary education (12.6% and 14.6%, respectively), it did not for those with a high-school education or women. González de la Rocha (2008) showed that the program improved occupations, mainly of the indigenous beneficiaries and that the youths who migrated from their localities of origin often have access to better employments. In addition, Ibarrarán and Villa (2010) concluded that the program did not have any impact on the quality of the youth’s employments. In the case of studies on intergenerational occupational mobility, there is a consensus that, after 10 years, PROSPERA had not achieved positive impacts (Rodríguez-Oreggia and Freije, 2008; Rodríguez-Oreggia and Freije, 2010; Yaschine, 2015). Furthermore, Yaschine (2015) identified higher rates of absolute mobility and greater social fluidity for women and youths who migrated, than for men and those who stayed in their locality of origin, respectively. Additionally, she documented the coexistence of social origin and education as important determinants of occupational attainment. Such results lead program officials to conclude that “after ten years of operation in rural areas, that observed impacts for youths, regarding social mobility, occupation and income were lower than expected” (SEDESOL and Oportunidades, 2014: 2). These results were used as justification for the creation of the labor inclusion and productivity sub-components of the recent linkage component (e.g. Componente de Vinculación). In the framework of this area, PROSPERA has sought to promote the formal labor insertion of the beneficiaries in urban areas and the productive 8 activities of the beneficiaries in rural areas, through the coordination with the supply of other federal programs on the subject. Notwithstanding, the effects of this line of work have been limited, among other factors, because of difficulties in the interinstitutional coordination, as well as because of the reduced coverage of and budget for labor and productivity programs (Yaschine, 2018). More recent studies that analyzed a longer period of exposure to the program (Kugler and Rojas, 2018; Parker and Vogl, 2018) documented impacts of greater magnitude on the indicators of labor insertion of rural youths (labor participation, hours worked, labor income and transmission from informal to formal employment), although these studies did not estimate the impact on intergenerational occupational mobility.12 Based on this research background, it has become relevant to analyze intergenerational occupational mobility 20 years after the introduction of PROSPERA and to include a cohort of youths that have reached a more advanced age and whom may experience more definitive results in labor outcomes. It is reasonable to suggest that it is possible to obtain more favorable results in this long-term study, compared to those performed a decade ago, although, as is detailed in section IV, the characteristics of the sample limit the scope of the analysis. III. Conceptual framework of the studies on intergenerational social mobility The technical approach we used is based on other intergenerational social mobility studies, which focus primarily on analyzing the degree and characteristics of the association between the social origins of individuals and their destinations.13 Its emphasis lies in investigating the existing level of equality of opportunities, as well as the different factors and mechanisms explaining the intergenerational transmission of socioeconomic advantages and disadvantages in societies (Ganzeboom, Treiman and Ultee, 1991). In research, this translates into the analysis of the association between the ascribed characteristics and the attainments in different dimensions of welfare (Ganzeboom, Treiman and Ultee, 1991). In sociology, studies on intergenerational mobility have focused their analysis primarily on the occupational area, in accordance with the argument that the division of labor is one of the central axes of social stratification and social inequality (Ganzeboom and Treiman, 1996) and that labor compensation largely determines the economic opportunities of individuals and, therefore, their access to well-being (Hauser and Warren, 2001). Thus, the occupational structure is seen as a summary of the structure of distribution of resources and opportunities in a society (Solís, 2005). The research on intergenerational occupational mobility enables, through its emphasis on labor, an approach to analyze the transmission of socioeconomic status between generations.                                                              12 Acevedo, Ortega and Székely (2018) found a positive correlation between exposure to the program and the probability of being employed, being formally employed and having a higher salary, although they make no estimations on the impact. 13 The social origins refer to a person’s conditions and circumstances of their early life and the closest point in time is envisaged as his or her social destination (Hout, 2015). 9 Intergenerational occupational mobility analyses have been developed around two analytical strands: the first is focused on the comparative analysis of patterns and rates of intergenerational occupational mobility, and the second is centered on the study of determinants of the attainment of individual status (Ganzeboom, Treiman and Ultee 1991; Erikson and Goldthorpe 1992; Treiman and Ganzeboom 1998; Erikson and Goldthorpe, 2001; Breen 2004a). These strands approach the study of social mobility from different perspectives, the first with the intention to describe the phenomenon at a macro-social level and, the second, with the aim to explain the micro-social stratification process behind the mobility system of society. In this study we applied both strands to answer our research questions. The analysis within the first strand is focused on the study of absolute mobility and relative mobility. Absolute mobility refers, in the case of occupational mobility, to the change of occupational position between generations which is influenced by structural change (economic development, productive changes and in the structure of labor positions, and demographic change), as well as by relative mobility. The latter refers to the pattern and strength of the association between origins and occupational destinations of individuals, regardless of the effect of structural change. Relative mobility is equated with social fluidity and equality of opportunities, as it compares the opportunities of mobility between groups with different social origins, that is, it indicates the level of inequality in the competition for opportunities offered in the labor market (Breen, 2004a; Breen, 2004b; Cortés and Escobar, 2007). Within this framework, the analyses have been performed both from an approach studying social mobility in relation to the movement of individuals in a vertical scale of occupations, as well as from another approach which visualizes the same as a movement between social classes (Hout and DiPrete, 2006). Those who argue for the first approach propose, based on empirical evidence, that the occupational structure follows a continuous hierarchical scale with differences in degree, which are not qualitative (Hout and DiPrete, 2006). The continuous occupational scale most frequently used in international studies on social mobility is the Socio-economic Index of Occupational Status (ISEI) developed by Ganzeboom, De Graaf and Treiman (1992), which measures the socioeconomic status associated to each occupation. Meanwhile, those who adhere to the second approach, the approach of classes, argue that social stratification may be represented by means of categorical grouping, which does not necessarily have a vertical structure but rather, mainly, qualitative differences. They consider that this perspective makes possible the analysis of individual mobility within the structural context and, thus, mobility may be seen as a process of mediation between the micro and macro-social levels. Various social classes schemes, seeking to represent the stratification in society and which result from different theoretical perspectives, have been proposed. The CASMIN scheme (Erikson and Goldthorpe, 1992),14 which has been applied in most contemporary studies on social mobility at an international level, seeks to capture the differentiation in the labor markets and production units related to the distribution of life opportunities (Hauser and Warren, 2001; Breen,                                                              14 The acronym refers to its creation within the project Comparative Analysis of Social Mobility in Industrial Nations. 10 2005). According to Sorensen (2001: 288), the CASMIN scheme, although following the theoretical principles and identifying qualitative differences, may also be associated with what may be called “stratum concept of class”, in that it groups individuals into strata according to the access they have to life opportunities or to a certain socioeconomic level derived from their position in the labor market. In this research, as will be detailed later, we use both the ISEI, as well as a classification of socioeconomic strata based on the CASMIN scheme. The second strand of the mobility analysis, introduced by Blau and Duncan (2001) at the end of the 1960s, is focused on the analysis of occupational status attainment. It differentiates from the previous strand focusing on analysis at the individual level (rather than the analysis of the mobility system in society as a whole) and by its multivariate study of the stratification process that leads to individual attainment (Goldthorpe, 2005). Blau and Duncan (2001) proposed a model that analyzes the influence of origin (occupation and education of the father) on occupational attainment mediated by non-ascribed factors (education of the child and occupational status of the child in their first employment). This model allowed the differentiation and comparison of the effect of the ascribed (the social origin) and non-ascribed factors, with special emphasis on the effect of education on occupational destination (Ganzeboom, Treiman and Ultee, 1991; Blau and Duncan, 2001). This strand of analysis was replicated by different researchers at an international level (Hout and DiPrete, 2006) and the proposed model has been the object of criticisms due to its over- simplification of the stratification process (Kerckhoff, 1995). In this sense, there have been efforts to account for additional factors which provide a more precise explanation of the mechanisms involved in the determination of the occupational attainment: Sewell, Haller and Portes (2001) and Sewell, Heller and Ohlendorf (1970) added psychosocial variables (mental or cognitive abilities, educational aspirations, influence of persons who are close), Kerckhoff (1995) argued in favor of the incorporation of institutional factors related to the educational institutions and the labor market and Lin (1999) and Kerbo (2006) refer to the studies that included variables related to social resources (social networks or social capital). Empirical studies have identified factors that influence the association between occupational origins and destinations. It has been shown that the ascribed factors (sex, race, ethnic background and family characteristics of origin) are strong determinants of both educational and occupational attainment (Blau and Duncan, 2001; Hout and DiPrete, 2006). Furthermore, education is highlighted as the most important factor of occupational attainment, with the potential to equalize opportunities between the individuals; although it may also operate in ways that favor intergenerational inheritance. In this sense, it is argued that the characteristics of the educational system (how stratified, specialized-differentiated and decentralized it is and its degree of linkage with the labor market) influence educational attainment and labor results (Kerckhoff, 1995; Shavit and Müller, 1998) and that, due to the effect of the unexpected educational expansion, the differences related to the quality in education may have replaced the quantity of education in importance when assessing the labor market (Breen and Jonsson 2005, Kerbo 2006). 11 There is a consensus that a trend towards the reduction of association between educational attainment and occupational destination has been observed (Breen and Luijkx, 2004; Goldthorpe, 2005), which may be due to the role of education as a “positional good”. That is, “what matters, regarding the returns to employment, is not the amount of education individuals have, but the amount relative to their competitors in the labor market” (Hirsch quoted by Goldthorpe, 1996: 494). This demonstrates that there is not a simple relationship between more education and more social mobility. In addition to individual and family traits, institutional factors clearly stand out as being crucial to the understanding of the intergenerational processes (Hout and DiPrete, 2006). Labor markets are one of the institutions that most affect stratification and social mobility, since they define the labor opportunities that individuals have (or do not have) access to. Furthermore, the role of public policies that favor (or hinder) equality of opportunities is emphasized. Through the creation of institutions, the regulatory framework and actions of specific policies, states may influence two of the processes that determine equality of opportunities: access for individuals to opportunities and resources, as well as the returns they can obtain for themselves in the market (Breen and Luijkx, 2004). A relationship between the welfare states with the existence of more fluent societies has been found, which leads to argue that the states that carry out policies explicitly seeking to counteract inequality achieve to protect the population from poverty and negative events (such as unemployment or retirement, for example) which may create downward mobility (Beller and Hout, 2006; Breen and Luijkx, 2004).15 The analysis of social mobility and equality of opportunities becomes greatly important when studying chronic poverty, that is, long-term poverty commonly persistent between two or more generations. In these cases, it is possible to refer to an intergenerational transmission of poverty, in which parents pass on to their children a set of features that increase the probability of those children experiencing poverty (Moore, 2005). This is an expression of social immobility; insomuch as individuals reproduce the socioeconomic characteristics of the previous generation. This immobility is related to the transmission (or lack thereof) of material, social, human, cultural and natural resources from parents to children and their development in contexts characterized by a structure of inequality of opportunities that limit their access to goods, services and opportunities which are indispensable to achieving higher levels of welfare (Sen, 1981; Kaztman and Filgueira, 1999; Moore, 2001; CPRC, 2004; Moore, 2005; Gough, McGregor and Camfield, 2006; Bird, 2007; Newton, 2007). With this as the backdrop, PROSPERA is analyzed as an example of a public policy with the State seeking to intervene to reduce the weight that social origins have on children and youths born in households with extreme poverty conditions and, hence, favor intergenerational mobility of these individuals, through actions that promote equality of opportunities.                                                              15 There is evidence that social fluidity is greater in socialist countries and in those with social democratic welfare states, such as the Scandinavian countries (Erikson and Goldthorpe 1992; Breen and Luijkx 2004; Beller and Hout 2006; Sorensen 2006). 12 IV. Methodological design This section describes the sources of information and the working sample, the characteristics of the study’s population sample, the design of the impact evaluation of the program, as well as the methods and variables used in the analysis of intergenerational occupational mobility and the determinants of occupational attainment. A. Sources of information and working samples The central source of information is the database from the Evaluation of Rural Households Survey (ENCEL in Spanish) 1997-2017 of PROSPERA.16 In addition to the information from ENCEL, the National Income and Expenditure Household Survey (ENIGH in Spanish) 2016, representing the makeup of the population at both the national and federal state levels, was used as a source of information to validate the socioeconomic strata, as well as compare the characteristics of our study group with the same age group from the population of the country. For the analysis on intergenerational mobility and determinants of occupational attainment we used the baseline from the ENCEL, provided by the Socioeconomic Characteristics of the Households Survey (ENCASEH in Spanish) collected in 1997, as well as the 2017 round of ENCEL. In the construction of the comparison groups we used information of the whole panel of ENCEL, as well as historic administrative data of PROSPERA that, among other variables, details the period and amount of cash transfers received between 1997 and 2017 for each household of the selected youths. Finally, for the estimation of the weights of the propensity score method we used the ENCASEH 1997.17 Between 1997 and 2017, ten rounds of the ENCEL have been collected: including seven rounds between 1997 and 2000 and also in 2003, 2007 and 2017. The rounds between 1997 and 2000, captured information from the entirety of the households residing in an experimental sample of 506 rural localities (320 treatment and 186 control) of high or very high marginality, located in seven states of the country: Guerrero, Hidalgo, Michoacán, Puebla, Queretaro, San Luis Potosí and Veracruz. However, the experimental design was lost starting in 2000 when the control localities were incorporated to the program. This has presented a methodological challenge for the impact evaluation, which has had to use quasi-experimental methodology.18 The last round, collected in 2017, prioritized information that would enable the long-term evaluation of PROSPERA, essentially to answer questions about its effects on children enrolled in the program between 1997 and 2000. For this purpose, 15,457 youths that in 2017 were                                                              16 The panel database comprises of variables from all rounds of ENCEL. See World Bank (2018b). 17 In some cases, where necessary information of ENCASEH 1997 was not available, the information from the closest round was used. 18 See INSP (2007) and Yaschine, Hernández and Urquieta (2008), for a discussion on the loss of the experimental design between 1997 and 2007.   13 between 15 and 35 years of age, and who comprise a subset of the sample of the panel of ENCEL, were selected.19 ENCEL 2017 is composed of four data-bases: 1) Household – which comprises information on the characteristics of the selected youth’s household, reported by a proper household informant; 2) Members – which includes information on the characteristics of all the members of the selected youth’s household, reported by a proper household informant;20 3) Youths – which comprises more detailed information on the selected youths, provided by them directly; 4) Providers – which includes more detailed information on the characteristics of the selected youths’ providers when the youths were 14 years old, reported by the providers directly.21 The implementation of the survey faced various limitations. One of the most important limitations was missing data, that is, a significant number of the selected youths could not be found. The survey was not designed to have a follow-up mechanism for the youths that have migrated from their localities of origin and only those residing in the same municipality, in some operational routes or in the cities of Mexico, Guadalajara and Monterrey, were able to be sought after. Therefore, data were recovered on only some of the migrant youths within the sample, particularly those residing within Mexico and in some of the aforementioned places. Hence, from the 15,457 initially selected, information on the households or members of only 12,470 youths was collected, and only 6,186 of them could be directly interviewed through a youths’ questionnaire. Furthermore, only in the case of 4,962 of these youths was it possible to directly interview their provider.22 In accordance with the interests of the research, our analysis required focusing on the youths that cover the following criteria:  Belonged to a household that in 1997 was eligible or that through the period of the panel had received benefits from PROSPERA.  Were between 18 and 35 years of age. The range starts at the legal age to enroll in high school and concludes at the maximum age of the selected sample for ENCEL 2017.  Were employed at the time of the study, since a study like this necessarily requires that individuals under study have entered the labor market.                                                              19 See World Bank (2018a and 2018c) for a description of the design of ENCEL 2017 and of the collected information. 20 In the household and members databases, new households were included (not interviewed in the baseline), which were created in recent years by some selected youths (World Bank, 2018a). 21 The age of 14 years old is commonly used by intergenerational mobility studies to define characteristics of social origin, because it is considered to possibly represent an important moment in the formation of the individual. 22 According to the World Bank (2018d), youths who were interviewed directly in the survey, in contrast with those who were in the sampling frame but not interviewed, come from households that, on average, are reasonably similar. This, in terms of age and schooling of the head of the household, as well as in the proportion in which the latter were employed as day laborers or workmen, the access to social security through the employment of the head of the household, the percentile of the poverty index and the size of the household. Nonetheless, it is possible to differentiate them by the lower degree of marginalization of the communities of origin exhibited by those who were interviewed, even though these differences were small in magnitude (World Bank, 2018d).   14  Have information on the characteristics of their occupation and of their provider (from the data base of youths and providers).23 According to these criteria, we have prepared a working sample for each of the analyses of our study, which consider the individuals inserted in the labor market as our interest group. For the analysis of intergenerational occupational mobility, we had 3,084 cases, that is, those for which the variables of occupational stratum necessary for this study could be constructed, both for the youth and their provider. The sample for the study of the determinants of occupational attainment comprised of 3,310 cases, which had information available to construct the ISEI of their current occupation, which was the dependent variable in that analysis.24 The analyses, in each case, were performed both for the sample as a whole, as well as for the subgroups created by differences in sex, ethnic background and migratory status. Furthermore, the analysis of occupational mobility was performed by comparing groups which received different intensity of PROSPERA’s treatment. Additionally, different analyses by age groups (18- 27 and 28-35) were performed, as well as by type of household (which differentiates between those who live in their household of origin and those who live in a new household). Nonetheless, we decided not to include these results in the document due to the fact that: 1) we did not find differences between age groups in any of our analyses, and 2) we identified that the occupational differences between youths with different types of household actually reflect the occupational changes derived from migration experienced by many youths living in new households.25 It is important to note that the studies of intergenerational mobility usually have, as subjects of analysis, individuals who are in adulthood and have become autonomous agents with respect to their household of origin (CEEY, 2013). In the case of studies on occupational mobility, individuals that are around 30 years old are considered to have reached occupational adulthood and have achieved a stable occupation and labor position (Torche, 2010). In the case of our analysis, the study group is comprised, largely, by persons who have not yet reached such maturity, and by only a small fraction are close to or have exceeded 30 years of age. This suggests that the occupational results of this group may improve in the future.26 Appendix A describes, in greater detail, the process and criteria for the selection of our working samples. Furthermore, descriptive data are presented in this appendix which provide evidence that some of their sociodemographic, educational and labor characteristics in 2017 are similar to those of the group of youths originally selected in ENCEL 2017, which mitigates possible biases in the study sample relative to the one originally selected. The exception to this is the fact that, given the emphasis of the study is analyzing youths who are employed, the lower                                                              23 In the cases where the questionnaire of the youth, but not of the provider, was available, relevant information on the provider was obtained through the questionnaire of the youth. 24 This second working sample adds 226 cases to the previous. The difference is that for the analysis of the determinants of occupational attainment it is possible to have missing data for the provider. 25 We consider it is preferable to present the analysis differentiating the migratory status instead of type of household. 26 In principle, an improvement in the occupational characteristics of the youths could be expected as they approach occupational maturity, which would suggest that the results of this study may be an underestimate. Nonetheless, as referred to in the previous footnote, the empirical analyses performed with two age subgroups within our study group do not show more favorable results for those who are older. 15 rate of female labor participation implied that our working samples include a lower percentage of women. As previously suggested, the sample selection of ENCEL 2017 and the restrictions derived from survey collection imply limitations to the representativeness of the results of this study. Therefore, despite the similarities between our working samples and the originally selected sample, we consider that the results of this study are valid only for the subpopulation analyzed in this study and may only be considered as indicative (but not representative) of the beneficiary population of PROSPERA in rural areas. Furthermore, it is reasonable to believe that the results may underestimate the characteristics of occupational mobility of youths, due to the fact that the analysis did not include all of those individuals from the original sample of ENCEL who have migrated from their locality of origin.27 B. Characteristics of the study group Our first approach to characterization is to compare our study sample to similar youths at a national level. Table 1 shows some of the educational and labor characteristics of our study group (ENCEL 2017),28 as well as of occupied youths in the same age range at the national level and in rural areas of the country (ENIGH 2016). For context, to interpret these data, it is important to emphasize that, while the sample of ENIGH 2016 (national and rural) includes youths that reside in localities of all degrees of marginality, the youths in our sample currently live predominately in localities with high and very high levels of marginalization (8.5% and 72.6%, respectively), while only the remaining 18.9% live in localities with a lower degree of marginalization.29 When comparing the characteristics of both groups, there are some observed differences between the two that point towards a more unfavorable situation for youths in our study’s sample, both in the area of education, as well as in their employment and labor outcomes. In comparison with the reference populations, in our sample, a lower percentage of youths are still studying, they have fewer years of average schooling, there is a greater proportion of them who are self- employed, they have lower average labor income (their labor income represents 60% of that received by an average youth at the national level), they are working in occupations with lower status and a smaller proportion of them has access to health services. These contrasts are less                                                              27 As previously mentioned, Yaschine (2015), with data from ENCEL 2007, which included a representative sample of migrants, identified higher rates of absolute mobility and higher social fluidity for this group. 28 The working sample for the study of the determinants of occupational attainment is used since it presents a greater number of cases. 29 The youths in our group come from rural localities that in 1995 had a high (62.4%) or very high (37.6%) degree of marginalization. As can be observed, a smaller portion of the localities of current residence has a high or very high degree of marginalization, which is due to the modifications in the marginalization level of the localities of origin where around 68% of our study group still resides, as well as to the change of residence to localities with lesser marginalization by part of the rest of the youths. The marginalization classification from the National Population Council (CONAPO, in Spanish) was used. For the localities of origin, the degree of marginalization for 1995 is used and for the localities of current residence the degree of 2010 was taken.   16 pronounced when compared to just the rural youths in the country, but they increase when compared to both urban and rural youths at the national level. In addition to comparing the characteristics of our study’s sample with the reference populations, it is also important to know the differences between the different subgroups comprising our sample. In table 2 some of the educational and labor characteristics of the youths are presented, disaggregating into subgroups of different sexes, ethnic backgrounds, migratory statuses and intensities of PROSPERA’s treatment received.30 A comparison of sexes suggests that females have better educational characteristics (one more year of schooling and a higher school attendance rate), however, despite possessing occupations with higher status, their labor conditions are more precarious (they have labor income that is 20% lower, a higher proportion them are self-employed and without pay, and they have lower access to health services), which is consistent with the existing knowledge on gender segregation in the labor market and gender inequalities in labor remunerations.                                                              30 The work sample for the study of the occupational attainment determinants was used since it presents a larger number of cases. 17 Table 1. Educational and labor characteristics of occupied youths between 18-35 years of age, ENIGH 2016 and ENCEL 2017 ENIGH 2016 ENIGH 2016 ENCEL 2017 Variables of interest Rural and urban Rural (n=3,310) Average age 26.84 26.45 25.98 Females (%) 40.79 37.49 38.55 Single (%) 42.07 36.83 38.88 Average years of schooling 10.85 9.00 9.27 Studies and works (%) 9.44 5.29 3.66 Only works (%) 90.56 94.71 96.34 Labor position (%) Employers 3.86 6.22 0.85 Wage-earning 83.14 69.16 79.91 Self-employed 8.23 13.28 15.85 Without pay 4.77 11.33 3.39 Average labor income 5159.71 4703.50 3405.68 Average ISEI 34.30 23.92 23.70 With health services (%)* 40.92 19.40 16.49 * Includes medical services of IMSS, ISSSTE, Local ISSSTE, PEMEX, Ministry of Defense, Ministry of Navy, universities and private. Source: Own elaboration from ENIGH 2016 and ENCEL 2017. When comparing different ethnic backgrounds, the educational and labor characteristics of both subgroups were similar, with the exception being that, although possessing occupations of similar status, indigenous individuals seem to receive lower labor income and a lesser proportion have access to health services, which also indicates an inequality in remunerations. It is possible that, in addition to discrimination due to their ethnic background, the disadvantages of indigenous individuals are associated with their context. Indigenous individuals not only come disproportionately from localities of very high marginalization (89.5% in comparison to 52.8% of non-indigenous), but also their current localities of residence are, in greater proportion, of very high marginalization (23.9% in comparison to 3.1% of non-indigenous).31                                                              31 The marginalization of the localities of origin data correspond to those from CONAPO in 1997 and the localities of current residence to those from 2010. 18 Table 2. Educational and labor characteristics of occupied youths between 18-35 years of age of ENCEL 2017, by sex, ethnic background, migratory status and intensity of PROSPERA’s treatment ENCEL 2017 (n=3,310) Low High Non- Non- treatment treatment Total Females Males Indigenous indigenous Migrant migrant intensity intensity Variables of interest (n=3,310) (n=1,276) (n=2,034) (n=865) (n=2,445) (n=760) (n=2,267) (n= 1,623) (n= 1,683) Average age 25.98 26.04 25.94 26.22 25.90 26.22 25.83 25.34 26.61 Single (%) 38.88 39.98 38.17 37.31 39.42 26.58 45.35 42.21 35.65 Years of schooling 9.27 9.73 8.98 9.13 9.31 10.02 8.99 9.56 8.99 Studies and works (%) 3.66 5.58 2.46 3.13 3.85 4.87 2.96 5.61 1.78 Only works (%) 96.34 94.42 97.49 96.87 96.11 95.13 97.04 94.39 98.16 Age at first employment 15.78 16.89 15.07 15.82 15.76 16.51 15.53 15.9 15.65 Labor position (%) Employers 0.85 1.02 0.74 0.70 0.90 1.19 0.75 0.68 1.01 Wage-earning 79.91 75.10 82.88 78.40 80.41 81.90 79.39 80.66 79.14 Self-employed 15.85 19.70 13.47 16.96 15.48 15.19 15.83 15.70 16.03 Without pay 3.39 4.18 2.91 3.95 3.20 1.72 4.02 2.97 3.81 Monthly labor income 3405.68 2929.30 3690.26 2991.29 3539.56 4344.84 3060.23 3400.62 3391.24 Average ISEI 23.70 27.12 21.56 23.02 23.94 27.90 21.89 24.40 23.01 With health services (%)* 16.49 14.43 17.78 12.92 17.75 30.65 11.17 17.67 15.34 * Includes medical services of IMSS, ISSSTE, Local ISSSTE, PEMEX, Ministry of Defense, Ministry of Navy, universities and private. Source: Own elaboration from ENCEL 2017. The more important contrasts can be observed when comparing youths who migrated from their localities of origin, with those who still reside in them. In comparison to non-migrants, the migrants presented better education and labor characteristics: a greater proportion attended school, have had an additional year of schooling, had a labor income that is 40% higher, have occupations of higher status, and the portion that have access to health services is three times greater. The more favorable labor characteristics observed with migrants may be associated with the change of context of labor insertion because the youths who migrated from their place of origin now settled predominantly in urban localities with lower levels of marginalization. Of the migrants in the study’s sample, 73.9% currently live in an urban locality and 49.5% in a metropolitan area, with the Mexico City Metropolitan Area being the greatest recipient, containing 20.5% of the migrants. Furthermore, only 28.2% of the migrants live in localities with high or very high level of marginalization, while 96.4% of the non-migrants do.32  In the case of the intensity of PROSPERA’s treatment subgroups, it appears that only the group with low treatment intensity includes a greater percentage of youths that are single and whom are currently studying in addition to working.33                                                              32 The marginalization data correspond to CONAPO from 2010. 33 In this case, for the treatment intensity data, the weights described in the following section are not applied. 19 As previously mentioned, unemployed youths are not included in our analysis. As shown in table 3, this group of youths compares similarly to the employed youth in average age and schooling, as well as in the proportion of indigenous population, of people who are illiterate and of those who still attend school. The main difference between both groups is the high percentage of females and lower percentage of people who are single in the unemployed group. Although the data of ENCEL 2017 did not include information on the activities carried out by these youths, due to the low school attendance rate in this group, it is reasonable to think that females are committed to household activities and it is not possible to know if they will later be incorporated into the labor market or not. Table 3. Characteristics of occupied and unemployed youths of ENCEL 2017 ENCEL 2017 Youths questionnaire (n=5,101)* Variables of interest Occupied (n=3,310) Unemployed (n=1,449) Average age 25.98 25.87 Females (%) 38.55 89.99 Indigenous (%) 26.13 25.53 Single (%) 38.88 22.22 Illiterate (%) 3.08 2.42 Attend school 3.66 4.89 Average years of schooling 9.27 9.34 * There are youths for whom it is not possible to know their condition of activity, because they did not answer the relevant questions. Source: Own elaboration from ENCEL 2017. C. Design for estimating the effects of PROSPERA Defining a design to estimate the impacts of PROSPERA represents a methodological challenge at the present time. One possibility would be to use the experimental design established at the beginning of the program. This design only enables us to examine differences between households within a period of 18 months in which they received intervention. After 20 years, this represents 7.5% more exposure to the intervention than the original treatment group. Apart from the difficulties in estimating and correcting for the effects of this sample loss, sampling inaccuracies and the probable interaction between treatment groups during the duration of the experiment (Faulkner, 2012), it is not completely clear how the measured effects, if any, from this small variation in the period of exposure at the end of the 1990s, should be interpreted. Resorting to quasi-experimental methods to estimate the effects of PROSPERA, however, is not an easy task. Throughout the past 20 years, the program has adapted not only to the demands for continuous improvement of its design and operational processes, but also to different emergencies and changing political climates it has experienced. Throughout this period of time, there have been numerous changes, including the creation and elimination of program components, as well as 20 modifications to all of the current components (Yaschine, forthcoming). This dynamic, typical of long-term interventions, hinders our ability to estimate the effect from PROSPERA’s intervention with the required precision. Even if only the more stable components in the history of the program (education, health and nutrition) are considered, it is not easy to estimate the effects of PROSPERA on occupational mobility and occupational attainment of the youths for various reasons. First, in the educational component, even if we focus attention exclusively on the accumulated amount of scholarships granted as a proxy for treatment intensity, in this specific case, self-selection would be possible, in so much that the transferred amounts reflect individual behavior. Ultimately, households with youths who are better adapted to school life, possess greater abilities or higher levels development, receive greater transfers due to showing lower rate of lagging behind in school or dropout.34 Second, the scholarships, although “labeled” as such, pass entirely straight into the transfer bag of the household as a whole, which translates into different levels of economic advantages depending on the size and composition of the household. Thus, in some respects, the whole household benefits from the educational component of the program’s transfers, and the degree of economic advantage provided by the component depends on its structure. Third, and closely related to the previous issue, since the program’s transfers are consumed by the household, and given the structure of households, it is practically impossible to separate the effects of the different components of the transfers. What is clear is that the manner in which the program intervenes in the lives of the youths depends on the number of individuals that make up the household and their respective ages, as well as the length of time they remain beneficiaries of the program. To overcome the difficulties associated with the evolution of the program’s intervention and the selectivity of its support package due to conditionalities of that package, instead of using a reasonably stable component or the volume of the transfers as a proxy for the intervention, in this study we propose to statistically evaluate the intensity of exposure of the households to the program, exploiting the longitudinal design of ENCEL. Specifically, our approach assumes, based on the operational rules of the program and a review of its administrative records, that the demographic trajectory of the household, and the temporal distribution of the transfers, are good indicators of the intensity of the program’s treatment to which the households have been exposed. In order to exploit this approach, in practice, we used Structural Equation Models (SEM),35 to estimate the treatment variable as a latent class variable (unobserved), indicated (measured) by                                                              34 This issue has been discussed for the case of Oportunidades by Attanasio, Meghir and Schady (2010). As an outlet for this endogeneity of transfers, Kugler and Rojas (2018) choose to use the number of years that each individual was eligible to receive the scholarships granted by PROSPERA. 35 For such purpose, we propose an approach that uses a latent class analysis (LCA, a specific case of SEM) (Goodman, 1974) for the estimation of the counterfactual distributions. The latent classes approach for causal analysis in experiments with random assignment was used by Frangakis and Rubin (2002) to control selectivity caused by non- compliance after assignment.   21 the historic record of a set of variables related to the composition of the household and to the cash transfers it received. We have constructed our treatment variable using seven available variables in the ENCEL panel:36 (1) the length of exposure for households (period in which the households appear as beneficiaries of the program), (2) the total amounts transferred into a household up until 2003, 2007 and 2017, (3) the number of people comprising of a household (excluding “temporary” members), (4) the number of youths in a household of primary school age, (5) the number of youths in a household of secondary school age, (6) the number of youths in a household of high school age, and (7) the number of elderly adults in a household. It is important to note that estimating the degree of exposure of the households to the program in this manner is similar, in spirit, to the strategies that use deterministic functions of exogenous variables.37 The main difference of this study with other efforts to estimate the effects of the program is that, instead of calculating years of eligibility according to the individual’s age or potential transfers in accordance with family composition, it is estimated empirically, taking advantage of ENCEL’s complete panel, allowing a better partitioning of the working sample in order to differentiate the types of households that are statistically homogeneous, according to their historic domestic arrangements and cash transfers received from the program.38 We consider that the empirical correlation between the households’ composition and the temporal distribution of the transfers hardly reflects the disposition of certain type of families with some parents being more concerned over their children than others, or children with having greater cognitive abilities or being better adapted to school life. Rather, correlation is a result of households’ observance of the operational rules of the program over time. Therefore, we take the resulting latent classes as groups of households with different intensities of exposure to PROSPERA. The results of the latent classes analysis strongly suggest that a model with two classes fits the data well, and that a model with three classes does not provide an additional improvement in fit with the data (see Appendix B). Based in this analysis, we have determined that the model of two classes is most accurate for our study. According to the results of our selected model, the association of the latent class variable is strong and meaningful.39 The class which was labeled as being low treatment intensity has consistently produced higher probabilities of having smaller household sizes, with fewer youths of school-age, fewer elderly adults, histories (7%) of shorter participation in the program and, correspondingly, a lower volume of cash transfers (34% per household and 10% per adult equivalent using a uni-parametric scale), in comparison with the other class. This class groups                                                              36 It is important to note that each of these variables is in the ENCEL panel with different periodicity and pattern of missing values. 37 See Kugler and Rojas (2018) and Fernald, Gertler and Neufeld (2010). 38 The evaluation of the potential transfers is a particularly complex task since there is no information in the panel between 2007 and 2017. 39 The details of the 293 parameters evaluated are not shown for brevity, but they are available to interested readers if requested from the authors of this study.   22 49% of the youths in our working sample. Its demographic trajectory is summarized in the left panel of chart 1. In the right panel of chart 1, the class which was labeled as high treatment intensity includes larger households, with more youths of school-age, more elderly adults, longer histories of participation in the program and superior amounts of cash transfers received from PROSPERA.40 Appendix B presents the descriptive statistics of each group. Certainly, there are good reasons to think that different structures of the households are systematically associated with different development opportunities. Although the larger households on average participate more in the program, it is also true that they may be exposed to different social risks. In this regard, for example, it is possible to find significant statistical differences between both groups (of high and low intensity of treatment) in the average years of schooling of the head of the household and the logarithm of income per adult equivalent, both of which are larger for smaller households (see the descriptive statistics in Appendix B). Chart 1. Demographic trajectories of the households by class of intensity of PROSPERA’s treatment   a. Low treatment intensity (49%)      b. High treatment intensity (51%) Note: For the latent classes analysis we used the working sample of the model for the determinants of occupational attainment, which has the highest number of cases (n=3,310). Source: Own elaboration from ENCEL 1997 and historical data of cash transfers of PROSPERA. Due to the systematic differences observed between the households receiving different intensities of PROSPERA’s treatment before entering the program, such as the above mentioned, it is necessary to resort to quasi-experimental methods in order to minimize the lack of comparability of our groups at the level of the analyzed youths. For that purpose, we have estimated sampling weights using the methodological extension to propensity score matching                                                              40 It would also be possible to further exploit, the external variation provided by randomization of the treatment and control groups originally selected in 1997 to evaluate the program. In principle, this strategy would enable us to add to the high intensity group an average of 18 more months of exposure. However, to follow this line of analysis would require losing half of our working samples (all households for the original treatment group that, according to their household structure, would be assigned to the low intensity group in our classification) and therefore we would be losing statistical power.   23 proposed by Imai and Ratkovic (2014).41 This strategy enables the analysis of a pseudo-sample of youths in which the characteristics of their households of origin, prior to the intervention of the program, are not associated to the different intervention levels to which they were exposed (Ridgeway et al. 2015).42 Appendix B shows the variables used in the construction of the weights and the favorable balance achieved through their application. The weights obtained were used in the construction of the mobility tables for the groups with low and high treatment intensity, and for the purpose of estimating the effect of the program on both absolute occupational mobility (through the mobility rates) and on relative mobility (through the use of log-linear models). D. Methods and variables for the study of intergenerational occupational mobility This section describes the methods and variables used to answer the first set of questions posed. This requires the analysis of both the absolute mobility and the relative mobility of the youths of the sample and the relevant subgroups. First, we present the scheme of the occupational strata used, followed by the mobility tables and log-linear models. D.1 Occupational strata scheme For the analysis of intergenerational occupational mobility, we used an approach to occupational stratification that seeks to establish a relation between intergenerational mobility and the improvement of the socioeconomic status of the beneficiary youths of PROSPERA. This scheme was used by Yaschine (2015) for the analysis of intergenerational occupational mobility with data of ENCEL 2007 and is based on the scheme proposed by Solís and Cortés (2009) for Mexico, which, in turn, used elements of the CASMIN classification.43 An ordinal scheme is proposed, which assumes that a position in the highest occupational stratum grants greater access to life opportunities and, therefore, allows the identification of upward and downward movements of the youths. In table 4 our approach to occupational stratification is presented, which is ordered                                                              41 In methodological literature, the causal inference methods based on matching are often introduced as (1) a quasi- experimental technique to contrast sample treatment and control cases, or otherwise as (2) a non-parametric method of fit for systematic patterns of assignment to treatment when it is not reasonable to rely on the estimators of a simple parametric regression (see Morgan and Winship, 2015). 42 The estimations based on inverse probability weights, introduced by Rubin (1985) and extended to the marginal structural models in Robins et. al (2000), have been extensively used to control the selection bias in treatment variables. 43 The Solís and Cortés classification (2009) was the first in the country validated with representative data at the national level, using the ENIGH 2004. Yaschine (2015) used and adapted this classification for the analysis of the ENCEL 2007 data.   24 from the highest to the lowest stratum, and describes, in general terms, the occupations included in each stratum.44 Table 4. Occupational strata scheme Stratum Description Professionals, officials and directors of the public, private and social sector; technicians, educational workers, art, workers in entertainment and Non Manual (NM) sports/athletics industries, department managers, coordinators and supervisors of administrative activities and services; support workers in administrative activities Employers and employees of established commerce Commerce Manual work employer (non-farmers); bosses, supervisors and other control workers in artisanal and industrial manufacturing and in repair and maintenance activities (except for construction workers); operators of fixed machinery in continuous movement and equipment in the industrial High skilled manual (HM) manufacturing process (except for operators of specialized mobile equipment for construction); drivers and assistant drivers of mobile machinery or transport vehicles (except for transportation powered by human or animal traction); firemen, policemen and workers of the armed forces. Assistants, day laborers and similar occupations in the artisanal and Low skilled manual industrial manufacturing process or in repair and maintenance activities; handcraft and industrial construction, installation, finishing and maintenance workers for buildings manufacture (LMM) and other construction activities; operators of specialized mobile equipment for construction. Street vendors and street service workers; personal services workers in Low skilled manual establishments; domestic service workers; watchmen and guards, drivers services (LMS) of transportation powered by means of human or animal traction. Workers in the agricultural sector Agriculture Source: Own elaboration. Table 5 presents the total income, average schooling and ISEI average of individuals between 30 and 64 years of age belonging to each occupational stratum, according to the income data from the ENIGH 2016.45 Based on these data, it is possible to affirm that the proposed order is related                                                              44 Occupational classifications commonly assign individuals to an occupational stratum through the combination of three variables: occupation, position in the occupation and size of the establishment. Unfortunately, ENCEL 2017 only contained information on the occupation for youths and providers, so the assignment of occupational strata was based on this variable. This restrains the construction of the variable, however, in the case of this sector of the national population the variability of the other two variables is limited, since most are in a wage-earning position and work at small establishments. We consider that, in the absence of the other variables, occupation provides a proper approach to the relevant stratum. 45 As previously mentioned, this age range is considered to include adults with a stable occupation and job position.   25 to differentiated access to life opportunities which follows an ordinal logic, with non-manual workers at the top of the stratification and agricultural workers at the base.46 Table 5. Distribution, total income and average schooling and ISEI, by occupational stratum. Individuals between 30-64 years of age, ENIGH 2016. Total Schooling ISEI Stratum Total (%) Males (%) Females (%) income (average) (average) (monthly) NM 23.57 51.05 48.95 6,146.03 14.4 61.76 Commerce 11.56 45.05 54.95 4,194.14 9.93 37.16 HM 22.63 72.90 27.10 3,478.67 9.15 29.75 LMM 10.31 82.29 17.71 3,038.06 7.28 21.17 LMS 19.40 32.00 68.00 2,332.47 8.03 25.8 Agriculture 12.53 72.94 27.06 1,818.51  5.38 13.8 Source: Own elaboration from ENIGH 2016. In general, the data show a descending pattern of the total income, schooling and ISEI, with one exception. The LMM stratum clearly has higher income than the LMS, but their average schooling and ISEI are lower. This may be due to the fact that, as shown in table 5, these strata show differentiated composition of sexes, with the LMM stratum predominantly made up of males and the LMS of females. Despite having more average schooling, the more feminized stratum registers lower income, which is an example of gender inequalities in the labor market. Furthermore, in relation to the ISEI, the difference in order may be due to some occupations in Mexico not providing an income at a comparable level to that in other countries captured in the index (Solís, 2010). While the order of strata may be considered ordinal, as observed, the differences in income are not equidistant. That is, the movement between the different strata may mean changes of different magnitude in the average access to income and other life opportunities. Furthermore, it must be considered that a movement to an adjacent stratum (or nearby) represents a change of lesser magnitude than one that is of greater distance. D.2 Methods for the analysis of absolute mobility Absolute mobility analysis was based on calculations derived from tables of intergenerational occupational mobility, which are contingency tables in which each row corresponds to a different occupational origin (the stratum of the providers) and each column to an occupational destination                                                              46 The order remains when using labor income, except for the stratum of commerce, which has an average labor income that would place it in the same level as the HM stratum, due to fact that its total income has a higher composition of non-labor income than the other strata. We consider it is correct to use the total income because belonging to an occupational stratum is related not only to labor income, but also to access to other sources of income and life opportunities in general. 26 (the stratum of the youths). We used two measures of absolute occupational mobility. The first was mobility rates, which is the proportion or percentage of observations that experience immobility or different types of mobility (upward or downward, regardless of the number of positions they have moved) in relation to the total. The second was the output percentages, also known as the row distribution, which “records the distribution of destinations for each category of origin” (Hout 1983; 11), where the summation for each category of origin is 100%. These percentages may be interpreted as the probability that individuals have of attaining a certain occupational destination, given each individual’s origin, and considering both structural mobility and relative mobility. D.3 Methods for the analysis of relative mobility We used log-linear models to study relative occupational mobility. The log-linear models were fitted based on the contingency tables and allowed us to study the association or independence between the rows and columns of the tables, isolating the change of the table’s marginals. As previously mentioned, in our case, the contingency tables are occupational mobility tables, in which the rows are the occupational strata of the providers and the columns are the occupational strata of the youths. The application of the log-linear models for the study of intergenerational occupational mobility enabled us to isolate the structural change and measure the social fluidity or equality of occupational opportunities. The starting point of the log-linear models was a double-entry frequency table, with rows and columns. In the case of investigations on mobility, the rows refer to the origin (characteristics of the provider) and the columns to the destination (characteristics of the youth). It is also possible to carry out log-linear models derived from a triple-entry table (this is, the comparison of two mobility tables). The third entry refers to another relevant variable that may influence intergenerational mobility (in our case: sex, ethnicity, migratory status and treatment intensity).47 In order to identify the pattern and strength of the association between occupational origins and destinations we tested 11 different theoretical models that propose different specifications of the interaction parameters between the rows and the columns (for double-entry tables), as well as between the rows, columns and the third variable (for triple-entry tables). These theoretical models propose a hypothesis  on the origin-destination association, which may be governed by different processes such as, among others, intergenerational inheritance and barriers to long- distance mobility.48 The theoretical models are described in detail in Appendix C. As an example, we will briefly describe the models that are relevant in the analyses presented in section 5.2. The homogeneous diagonal (model 2) and the diverse diagonal (model 3) each parameterize intergenerational inheritance, with the difference being that the first case assumes a                                                              47 For the log-linear models, see Agresti (2007) and Xie (1992). 48 Sorokin (1974), identifies “circulation factors” that facilitate or prevent the transition from one social stratum to another, which may be related to the characteristics of the individual, the institutions, the political and economic system, etc. From a more pragmatic perspective, the social mobility barriers may be associated with the distance between the strata, where a greater distance implies greater obstacles to upward mobility (Solís, 2016). 27 constant inheritance for all occupational origins, while in the second case inheritance variates per stratum of origin. The trait that each of the proposed four corner models (models 8, 9, 10 and 11) have in common is that, in addition to parameterizing intergenerational inheritance (either as homogeneous or diverse diagonal), they include parameters that shape the long-distance mobility barriers of both the extremes of the occupational stratification (either as barriers with homogeneous or differentiated intensity). That is, they modeled the barriers in long-distance downward mobility from the stratum NM and in the long-distance upward mobility from the agricultural stratum. We applied the models in the double-entry analyses to study the relative mobility of the sample in its entirety, as well as in triple-entry analyses to test if there are differences in the relative occupational mobility between subgroups, specifically those differentiated by sex, ethnic background, migratory status and treatment intensity. In the case of the triple-entry analysis, we compared between log-linear models, which assume there is no interaction with a third variable (constant fluidity), and models that assume there is (non-constant fluidity) to identify whether relative mobility is different between each pair of the analyzed subgroups. For the triple-entry analysis, we also used the log-multiplicative model (Unidiff) to provide additional evidence on whether or not differences between the subgroups exist. The log- multiplicative model’s purpose is to test the interaction effect of the third entry when comparing two mobility tables.49 The Unidiff models generate a parameter that provides an account of the strength of the association with the third layer or variable. If ∑ 1 means there is no difference in the relative mobility of the two groups, if ∑ 1 means that the reference category has displayed greater relative mobility and a ∑ 1 means that the reference category has displayed lesser relative mobility. To evaluate the fit of the log-linear models and select those that better described the data, we used the Bayesian Index Criteria (BIC). The BIC, according to Raferty (1986), is a good selection criterion since it favors more parsimonious models, provides a better fit to the data and fits to the number of parameters and size of the sample.50 The BIC favors models with smaller values. E. Methods, variables and analytical model for the study of the determinants of the occupational attainment For the analysis of the determinants of occupational attainment we based our model on the model originally proposed by Blau and Duncan (2001 [1967]), which identifies the relationship between different factors associated with the social origin (occupation and education of the father) and non-ascribed factors (education and first occupation of the individual) in the process                                                              49 For the log-multiplicative model, see Xie (1992). 50 The or deviance. However, this criterion has a flaw of being sensitive to the size of the sample. The value of BIC penalizes by the degrees of freedom for the larger samples at a rate of log(n) and considers the degrees of freedom of the model. 28 that leads to individual occupational status. This model was empirically applied using path analysis. As mentioned above, we propose an adaptation to such model that includes additional determinants of occupational attainment that have been suggested in relevant literature. Specifically, we agree with the proposal of Sewell, Haller and Portes (2001 [1969]) that considers psychosocial variables, although we only aggregate cognitive abilities (from the provider and the youth), in the absence of information to build the other variables used (academic performance, educational and labor aspirations, influence of third parties).51 Furthermore, instead of following path analysis, we apply a model of structural equations, which incorporates latent variables. Figure 1 is a graphic representation of our proposed analytical model as a starting point of the analysis, which was subject to empirical tests that resulted in the selection of a more parsimonious model that will be presented in section VI. Figure 1. Structural model of the determinants of occupational attainment   Source: Own elaboration.                                                              51 We carried out the exercise of including a latent variable for social capital in the model. However, the manifest variable related to social capital in ENCEL 2017 was not adequate, neither theoretically, nor statistically. The latent variable had a bad fit, with a low level of composite reliability ( = 0.22). 29 This structural model includes five latent variables (provider’s cognitive abilities, as well as youth’s social origin, cognitive abilities, education and first occupation) and a manifest dependent variable (youth’s occupational attainment). Each latent variable is constructed from the manifest variables based on a confirmatory factor analysis (Brown, 2015). Each latent variable and the model’s response variable are described below. Appendix E will provide more details on the construction of latent variables (and on the manifest variables), as well as the calculation of the composite reliability .  The social origin is constructed from the manifest variables, such as the ISEI of the provider’s occupation, the schooling of the father and the mother when the youth was 14 years old and an index of assets and services in the household in 1997. The composite reliability is  = 0.60.  The youth’s and provider’s cognitive abilities are measured with manifest variables created by two cognitive tests that are affected by education: 1) Raven’s progressive matrices (Alderton and Larson, 1990; Bilker, et. al., 2012), that are used to measure inductive reasoning and are considered predictors of academic and labor performance, and 2) digit span, which is a test frequently used to measure short-term memory and attention span (Ostrosky-Solís and Lozano, 2006). The measurements of both youths and their providers were performed using data from 2017. Nonetheless, both are granted a place in the model preceding the current occupational attainment. This is because the cognitive abilities are essentially developed during childhood and are determined to a large extent by family socioeconomic status (Evans and Schamberg, 2009; Hackman, et. al., 2015). Therefore, it assumes that the measurements performed on youths reflect their cognitive state prior to the current labor insertion and the ones performed on the providers may closely reflect their abilities when the youths were 14 years old. The composite reliability is  = 0.46 and 0.52 for youths and their providers, respectively.  The latent variable, youth’s education, is measured based on four manifest variables: youth’s years of schooling and three variables that combine having attended a grade of the educational level and the quality of the school the youth attended at the secondary, high-school and superior levels. The inclusion of the variables for educational attainment and school quality acknowledge that education should not only be conceived exclusively as a meritocratic factor. A measurement that seeks to approximate the role of education as a mediator between the origin and the social destination is proposed, which may favor (or not) greater equality of opportunity. The composite reliability is  = 0.92.  The youth’s first occupation is measured with the age of their first occupation and the ISEI of that first occupation. The composite reliability is  = 0.45.  The dependent variable is the youth’s occupational attainment and is measured through the ISEI of the youth’s current occupation. The model established a time sequence that begins with the two latent variables that represent the ascribed factors linked to the youths’ socioeconomic origin: their social origin and their provider’s cognitive abilities. Based on these initial variables, the stratification process may 30 follow different routes to bring about the current occupational attainment. These paths may be directly or indirectly guided by the youth’s cognitive abilities, education and first employment, which may play a mediating role between the ascribed factors and the dependent variable. The weight of each variable on the current occupational attainment comprises the sum of all of its direct and indirect effects. The model of figure 1 proposes that the two initial exogenous variables, which refer to the characteristics of the providers and their household of origin, are correlated with each other (). The endogenous latent variables refer to the youth’s cognitive abilities, education and first occupation. It is suggested that the provider’s cognitive abilities directly influence the youth’s occupational attainment (β11), cognitive abilities (β1) and education (β10). Furthermore, they have indirect effects on the dependent variable through the youth’s cognitive abilities and education. In the case of social origin, this may directly affect the youth’s occupational attainment (β9), as well as the youth’s education (β3) and first occupation (β4), and indirectly influence occupational attainment through the direct effects of both mediating variables on the dependent variable. The youth’s education has a direct effect on the youth’s cognitive abilities (β2), first employment (β5) and current occupational attainment (β7), and indirectly influences the dependent variable by the youth’s cognitive abilities and first occupation. The youth’s first occupation directly influences occupational attainment (β8). It is possible that there are other paths that connect to the latent variables, however, but we did not include more paths in order to avoid over-parameterization. The structural model of figure 1 is a synthetic representation and only shows the six latent variables and the dependent variable (which is a manifest variable). Still, it should be understood that the latent variables are constructed from the manifest variables previously mentioned. This type of model does not present the problem of co-linearity that is typically observed in regression models (Bollen, 1989). In the application of the measurement error model we estimated, through the confirmatory factor analysis, the loads of the manifest variables over the latent variables, and the coefficients of each path. The model includes both categorical and continuous variables, which is why the weighted least squares estimation method (WLSMV) was used, implemented in the program MPLUS 7.11 (Muthen and Muthen 1998-2013). Additionally, the database to estimate the model presents missing values and, therefore, the full information maximum likelihood method (FIML; Arbuckle, 1996) was used, both so that there is no need to impute values, and because it is an efficient method (Vargas and Lorenz, 2017). We estimated the model for the study’s sample in its entirety and also exploring differences by sex, ethnic background and migratory status. V. Intergenerational occupational mobility This section describes the results of the intergenerational occupational mobility analysis. First, the results corresponding to absolute mobility are presented followed by those referring to relative mobility. Both cases are presented for the sample in its entirety and for the subgroups for sex, 31 ethnic background, migratory status and treatment intensity. It is important to remember that the treatment intensity data are the result of the application of the weights described in section 4.3. A. Results of absolute mobility Chart 2 shows the distribution of youths and their providers on the different occupational strata. In the case of youths, the distribution of sex, ethnic background, migratory status and intensity of PROSPERA’s treatment subgroups is also presented. This descriptive analysis provides the first approach to intergenerational changes in the occupational stratum. It can be seen that the distribution of the providers in the study group is concentrated, largely, in the agricultural stratum, while the youths, although having an important presence in that stratum, are redistributed towards other strata higher up in the hierarchy. The comparison of sex groups shows relevant differences; women have greater presence in the stratum in the highest hierarchy (NM and Commerce), as well as in the LMS, while males continue to have a significant presence in the agricultural stratum. The differences between the two ethnic background groups was not clearly noticeable, except for there being a slightly greater presence of indigenous people in the agricultural stratum. Chart 2. Distribution of providers and youths by occupational stratum. Youths from 18-35 of age of ENCEL 2017 and their providers. 2.4 2.9 6.6 8.6 8.8 9.7 9.2 9.9 10.7 12.4 14.7 7.7 7.2 9.0 10.8 11.5 11.9 12.3 12.3 10.0 14.9 18.2 19.6 12.1 8.9 13.4 13.3 14.2 14.5 15.1 13.0 OCCUPATION 17.7 10.6 STRATA 13.0 12.7 12.5 20.9 13.2 12.3 PERCENTAGE NM 3.9 5.3 11.0 14.9 Comm 16.2 15.9 10.4 16.2 15.6 HM LMM 68.2 33.2 LMS 48.2 Agricu 28.9 46.4 41.1 35.9 37.6 33.9 34.1 15.6 9.1 TO TA L TOTAL FEMA LE MALE I N D I GE N O U S NON M I GR A N T N O N M I GR A N T LOW H I GH I N D I GE N O U S TREA TMEN T TREA TMEN T I N TEN SITY INTENSITY PROVIDER YOUTH Source: Own elaboration from ENCEL 2017 (n=3,084). The comparison of migratory status particularly stands out. It is evident that young migrants, in a strong contrast with those who remain in their localities of origin, hardly ever carry out agricultural occupations and have placed themselves in occupations belonging to the strata higher in the hierarchy. Both groups have the same percentage of youths in the agricultural stratum and the group with higher intensity shows slightly higher proportions in the NM, LMM and LMS strata, but lower in commerce and HM. 32 The occupational mobility rates experienced by youths are presented in table 6, calculated based on the mobility tables, and they complement the description of intergenerational changes. The reviewing the rates enables us to affirm that, although around half of the youths experienced upward mobility, nearly 40% inherited their provider’s occupation and the rest move downward on the stratum. When observing the behavior of the analyzed subgroups, the difference between sexes stands out, with higher rates of upward mobility for women, in contrast to a greater immobility by men. But the subgroups with the largest gaps between them are the migrant and non-migrant subgroups; the former of which experience upward mobility in 7 of each 10 cases, while non-migrants experience this in only 4 of each 10 cases, and have high rates of immobility. The comparisons by ethnic backgrounds and treatment intensity show very limited differences.52 Table 6. Intergenerational occupational mobility rates. Youths from 18-35 years of age of ENCEL 2017. Total, by sex, ethnic background, migratory status and treatment intensity. By treatment By sex By ethnic background By migratory status intensity Non- Non- Low High Type of Total Females Males Indigenous indigenous Migrant migrant intensity intensity Mobility (n=3,084) (n=1170) (n=1914) (n=830) (n=2254) (n=702) (n=2124) (n= 1521) (n= 1,563) Upward 49.2 66.8 38.5 48.7 49.4 74.2 39.3 47.5 49.3 Mobility (%) Immobility 37.6 21.6 47.3 41.3 36.2 14.1 46.8 38.5 38.3 (%) Downward 13.2 11.5 14.3 10 14.4 11.7 13.9 14 12.4 Mobility (%) Source: Own elaboration from ENCEL 2017. An analysis of the output percentages for the entire sample, as shown in chart 7, and in Appendix F for the subgroups, provides additional findings. If we consider the complete sample, we observed that the probability of having a destination at the base of the stratification is greater for all strata of origin (between 17 and 43% of the cases), while the probability of having a destination at the top is low (between 9 and 11% of the cases), except for those who have a NM origin (26% of the cases).                                                              52 An analysis of the average difference for the ISEI of occupations of the provider-youth duo provides complementary results, aligned with those previously described. On average, youths have occupations with 6.8 more points of ISEI than their providers (p<0.001). For women this difference is 10.1 points (in comparison with 4.8 points for men), while migrants present a difference of 10.8 points (in contrast to 5.0 points of non-migrants). In both comparisons (by sex and migratory status), the difference between the compared subgroups is statically significant. Additionally, indigenous and non-indigenous persons have occupations with an ISEI of around 7 points more than their providers, and the difference between both subgroups is not statistically significant. In the case of the subgroups by treatment intensity, youths with high intensity have on average difference of 6.8 points regarding their provider (in comparison to 5.7 points of youths with low intensity); the difference between both subgroups is statistically significant, even if it is only one point. As described in Appendix E, the ISEI has a range of 11.0 to 88.7. The changes described here mean there is an improvement in the youths’ occupational status in relation to their providers, although, on average, the differences are limited and may be considered short-distance movements. 33 Table 7. Output percentage. Youths from 18-35 years of age of ENCEL 2017. Stratum of the son or daughter (n= 3,084) NM Commerce HM LMM LMS Agriculture Total NM 26 8.2 16.4 12.3 16.4 20.5 100 Commerce 9.1 18.2 15.9 19.3 15.9 21.6 100 Stratum of HM 8 11.8 30.4 13.5 19.8 16.5 100 the provider LMM 11 11 19.7 26.2 14.2 17.8 100 LMS 9.2 15.8 15.8 8.8 23.8 26.7 100 Agriculture 9.2 11.4 11.2 10.6 14.6 43 100 Total 9.7 11.9 14.2 12.5 15.9 35.9 100 Source: Own elaboration from ENCEL 2017. However, women and migrants are less likely to have an occupation in the agricultural stratum and more likely to have a NM occupation (for any stratum of origin), in comparison with the men and non-migrants, respectively. In the case of the comparison by ethnic background it is observed that indigenous youths are slightly more likely for to have an agricultural destination when their provider’s occupation was in the four lowest strata in the hierarchy, while they are more likely to inherit their provider's occupation if they have an NM origin. The comparisons groups by treatment intensity do not present substantial differences. B. Results of relative mobility In this section we present the results of relative mobility for the study’s entire sample, as well as for the differentiated subgroups by sex, migratory status and comparison group according to intensity of PROSPERA’s treatment. The results of the fit and the parameters of the log-linear models for the entire sample are presented in the text. The results of the fit of the log-linear models for the subgroups for sex, ethnic background and intensity of PROSPERA’s treatment are included in the text, while the parameters of these models and the log-multiplicative models are included in Appendix G. The results of the analysis by ethnic background are not presented in this document, since no differences between the subgroups were found.53 B.1. Relative mobility of the complete sample, by sex and migratory status Table 8 shows the results of the different log-linear models described in Appendix C, in the case of the entire sample of youths, that is, these are models based on the two-way mobility table. According to the values of the BIC, we may conclude that models 10 and 11 have the best fit. These models indicate that the main factors that determine the association between occupational                                                              53 These results are available per request of the interested party. 34 origins and destinations are occupational inheritance (with differentiated strength by stratum of origin) and barriers to long-distance mobility. Table 8. Adjustment of log-linear models. Youths from 18-35 years of age of ENCEL 2017. Type of model g.l. G2 X2 BIC Value of P 1. Independence 25 251.97 268.98 162.38 0.00 2. Homogeneous diagonal 24 46.76 47.27 -39.24 0.00 3. Diverse diagonal 22 39.68 39.83 -39.16 0.01 4. Symmetry (2p) with homogeneous diagonal 23 42.43 43.13 -39.99 0.01 5. Symmetry (3p) with homogeneous diagonal 22 42.42 43.15 -36.42 0.01 6. Symmetry (4p) with diverse diagonal 21 37.57 37.42 -37.69 0.01 7. Symmetry (5p) with diverse diagonal 20 37.55 37.42 -34.12 0.01 8. Corners (2p) with homogeneous diagonal 23 43.49 44.23 -38.93 0.01 9. Corners (3p) with homogeneous diagonal 22 40.13 41.20 -38.71 0.01 10. Corners (4p) with diverse diagonal 21 34.25 33.84 -41.00 0.03 11. Corners (5p) with diverse diagonal 20 28.97 28.78 -42.70 0.09 Note: g.1. = degree of freedom; G2 = Deviance; X2 = Pearson’s chi square; BIC = Bayesian Index Criteria. Source: Own elaboration from ENCEL 2017 (n=3,084). The parameters of both models are presented in table 9. For our interpretation, we will use model 11, which provides a more detailed explanation. According to the parameters of this model’s main diagonal, occupational inheritance is strongest for those who have an agricultural origin (dp6=2.92, p<0.001), followed by those with a NM origin (dp1=2.27, p=0.003) and, finally, those with an origin in the intermediate strata (dp23=1.80, p<0.001). Youths with an agricultural origin are 2.9 times more likely to inherit their provider’s occupation than to have a different occupation. In the case of those of NM origin and of the intermediate strata the probability is 2.27 and 1.80 times, respectively. We found no evidence of barriers to long-distance downward mobility from an NM origin (esq1=0.68, ls), but there are barriers to long-distance upward mobility from the agricultural stratum (esq2=1.32, p<0.001). Continuing the analysis, we studied whether relative mobility is different between males and females. To do this, we fitted the three-way log-linear models (with both constant fluidity, which assumes there are no differences between groups, and non-constant, which assumes there are) and the log-multiplicative models. Table 10 shows the results of log-linear models by sex. Our statistical findings do not allow us to conclude with certainty which is the better fit, because the differences in BIC between some of the fluidity models in their constant and non-constant versions are very small (between 1 and 5 points in the case of the models 2, 4, 8 and 9). 35 Table 9. Parameters of selected log-linear models. Youths from 18-35 years of age of ENCEL 2017. (a) Model 10. Corners (4p) with diverse (b) Model 11. Corners (5p) with diverse diagonal diagonal. Parameter OR Value of P Parameter OR Value of P dp1 dp1 2.61 <0.001 2.27 0.003 (NM inheritance) (NM inheritance) dp23 dp23 (HM, LMM, LMS 1.80 <0.001 (HM, LMM, LMS 1.80 <0.001 inheritance) inheritance) dp6 dp6 (Agriculture 2.74 <0.001 (Agriculture 2.92 <0.001 inheritance) inheritance) esq 1.21 0.019 esq1 0.68 0.174 esq2 1.32 0.002 Source: Own elaboration from ENCEL 2017 (n=3,084). The results of the log-multiplicative version of the models 2, 4, 8 and 9, which present values of the coefficient different from 1 (between 0.74 and 0.79) for these theoretical models, provide additional elements to define whether or not there are differences between both sexes. This leads us to conclude that there is a difference in relative mobility between females and males, although it is not very strong. Females, in comparison with males, experience greater occupational fluidity or equality in occupational opportunities. The four mentioned models propose that relative mobility of youths is characterized by intergenerational inheritance, which has the same strength regardless of the stratum of origin and, furthermore, they propose either the existence of short- distance mobility or barriers to long-distance mobility at both extremes of the stratification. The interpretation of differences in relative mobility between both sexes is based on the analysis of the parameters of the four non-constant fluidity models. These models consistently show that occupational inheritance is stronger for males than for females. If we consider model 9, we observe that males are 2.45 times more likely to inherit their provider’s occupation than to have a different occupation (dp=2.45, p<0.001), while this value is only 1.87 times for females (dp=1.87, p<0.001). Furthermore, this model suggests that males with an origin in the agricultural stratum experience a barrier to long-distance upward mobility (esq=1.51, p=0.002), and for females we found no evidence of this (esq=0.99, ls). 36 Table 10. Adjustment of log-linear models. Youths from 18-35 years of age of ENCEL 2017. Comparison by sex. Constant fluidity Non-constant fluidity Type of model g.l. G2 X2 BIC g.l. G2 X2 BIC 1. Independence 50 285.96 309.47 72.13 2. Homogeneous diagonal 49 80.38 81.36 -129.17 48 77.49 77.76 -127.79 3. Diverse diagonal 47 72.34 71.55 -128.66 44 68.86 67.81 -119.31 4. Symmetry (2p) with homogeneous diagonal 48 76.65 77.65 -128.63 46 72.23 73.25 -124.49 5. Symmetry (3p) with homogeneous diagonal 47 76.64 77.67 -124.36 44 71.43 71.87 -116.74 6. Symmetry (4p) with diverse diagonal 47 70.92 69.99 -125.81 42 64.8 64.98 -114.82 7. Symmetry (5p) with diverse diagonal 45 70.86 69.94 -121.59 40 63.89 63.46 -107.17 8. Corners of (2p) with homogeneous diagonal 48 76.99 78.43 -128.29 46 71.62 73.49 -125.11 9. Corners of (3p) with homogeneous diagonal 47 74.87 77.1 -126.13 44 66.83 70.3 -121.34 10. Corners of (4p) with diverse diagonal 46 67.19 65.94 -129.54 42 61.21 60.57 -118.41 11. Corners of (5p) with diverse diagonal 45 63.95 62.92 -128.5 40 54.61 55.3 -116.46 Note: g.1. = degree of freedom; G2 = Deviance; X2 = Pearson’s chi square; BIC = Bayesian Index Criteria. Source: Own elaboration from ENCEL 2017 (n=3,084). Another relevant question is if, when the youths migrated from their localities of origin, they experienced different patterns and magnitudes of association between occupational origins and destinations. Table 11 shows the results of the models with and without constant fluidity, considering youths by migratory status. The non-constant fluidity models, which assume differences in relative mobility between both groups, are the ones with clearly a better fit, by presenting differences between 13 and 17 points of BIC regarding the constant fluidity models. Among the array of theoretical models, in the non-constant fluidity version, models 2, 4, 8 and 9, in that order, present the best fits. 37 Table 11. Results of log-linear models. Youths from 18-35 years of age of ENCEL 2017. Comparison by migratory status. Constant fluidity Non-constant fluidity 2 2 Type of model g.l. G X BIC g.l. G2 X2 BIC 1. Independence 50 297.51 327.44 83.67 2. Homogeneous diagonal 49 85.32 84.48 -124.23 48 63.41 60.26 -141.87 3. Diverse diagonal 47 76.6 76.1 -124.4 44 59.64 56.08 -128.53 4. Symmetry (2p) with 48 81.03 80.19 -124.25 46 57.08 53.65 -139.65 homogeneous diagonal 5. Symmetry (3p) with 47 81.03 80.2 -119.97 44 54.41 51.5 -133.76 homogeneous diagonal 6. Symmetry (4p) with diverse 46 74.67 73.6 -122.06 42 54.99 50.94 -124.63 diagonal 7. Symmetry (5p) with diverse 45 74.65 73.61 -117.8 40 52.16 48.66 -118.9 diagonal 8. Corners (2p) with homogeneous 48 82.12 81.18 -123.16 46 56.85 54.13 -139.87 diagonal 9. Corners (3p) with homogeneous 47 79.76 79.12 -121.24 44 53.94 51.62 -134.23 diagonal 10. Corners (4p) with diverse 46 70.69 69.34 -126.04 42 50.93 46.9 -128.69 diagonal 11. Corners (5p) with diverse 45 65.68 64.76 -126.77 40 46.45 42.71 -124.62 diagonal Note: g.1. = degree of freedom; G2 = Deviance; X2 = Pearson’s chi square; BIC = Bayesian Index Criteria. Source: Own elaboration from ENCEL 2017 (n=3,084). The results of the log-multiplicative models confirm that there are differences in relative mobility between youths who remained in their localities of origin and those who migrated. The values of the coefficient for these theoretical models are different from 1 (between 0.23 and 0.26), with non-migrant youths as a reference. That is, migrants experienced more relative mobility and the difference between both groups seems important. The parameters of the selected theoretical models present evidence that the difference lies mainly in the strength of intergenerational inheritance. In the case of model 2, which had the best fit, we observed that non-migrants have a 2.47 times higher probability to inherit their provider’s occupation than to have a different occupation (dp=2.47, p<0.001), while in the case of migrants this probability is only 1.26 times and its statistical significance is marginal (dp=1.26, p=.08). By incorporating additional parameters to the main diagonal in other theoretical models selected, the parameter of inheritance of migrants becomes statistically insignificant. The other three models show two more differences between youths who did not migrate and migrants. Model 4 indicates that non-migrants experienced short-distance mobility from all strata of origin (sim=1.22, p<0.001) and an interpretation of the parameters of models 8 and 9 suggests 38 that the barrier to long-distance mobility is stronger when their origin is in the agricultural stratum (esq2=1.35; p=.005). These parameters are not statistically significant for migrants. Hence, we may conclude that migrants experience more social fluidity or equality in opportunity than non- migrants, that is, those who migrate from their localities of origin achieve occupational position that break the intergenerational inheritance and the barriers in long-distance mobility. B.2. Relative mobility by intensity of PROSPERA’s treatment In order to identify if a higher intensity of PROSPERA’s treatment had an effect on the relative mobility of the youth beneficiaries of the program, the strength and pattern of association between occupational origins and destinations of both treatment intensity subgroups, which is associated to the size and structure of the households, were compared. It was expected that the households with higher treatment intensity would observe a greater relative mobility or equality of opportunity. Table 12 shows the results of the models with and without constant fluidity when youths were grouped by treatment intensity. As in the case with our comparison of sexes, the statistical diagnoses did not allow us to conclude with certainty which was the best fit since the differences in BIC between constant and non-constant versions of some of the fluidity models is very small (between 4 and 10 points in the case of models 2 and 3). When reviewing the results of the log-multiplicative version of these theoretical models, we saw that the values of the coefficient vary between 0.91 to 0.94, that is, they are close to the unity, but are under 1. This leads us to consider that there might be a difference in the relative mobility of both groups. But, if that is the case, the difference would be small. To delve into this line of analysis, we reviewed the parameters of the log-linear models 2 and 3 of non-constant fluidity. These models, in addition to having a small difference in BIC with regard to the corresponding constant model, have the better fit between the non-constant models. According to model 2, which does not differentiate between different intensities of occupational inheritance by stratum of origin, both groups are marked by a similarly strong inheritance (the group with lower treatment intensity has dp=2.09, p<0.001 and the one with higher intensity has dp=2.00, p<0.001). That is, according to this model, both groups have the same probability of inheriting their provider’s occupation. 39 Table 12. Results of log-linear models. Youths from 18-35 years of age of ENCEL 2017. Comparison by treatment intensity. Constant fluidity Non-constant fluidity 2 2 Type of model g.l. G X BIC g.l. G2 X2 BIC 1. Independence 50 339.35 336.47 125.52 2. Homogeneous diagonal 49 147.85 136.09 -61.71 48 147.66 136.36 -57.62 3. Diverse diagonal 47 133.21 119.4 -67.79 44 130.67 118.21 -57.51 4. Symmetry (2p) with 48 144.99 135.56 -60.29 46 144.76 136.06 -51.97 homogeneous diagonal 5. Symmetry (3p) with 47 144.39 134.36 -56.61 44 144.17 134.87 -44.01 homogeneous diagonal 6. Symmetry (4p) with diverse 46 132.6 119.54 -64.13 42 130.04 118.41 -49.58 diagonal 7. Symmetry (5p) with diverse 45 132.02 118.53 -60.43 40 129.48 117.48 -41.58 diagonal 8. Corners (2p) with 48 146.85 135.61 -58.43 46 146.65 135.92 -50.07 homogeneous diagonal 9. Corners (3p) with 47 143.96 133.27 -57.04 44 143.05 132.99 -45.12 homogeneous diagonal 10. Corners (4p) with diverse 46 130.03 116.39 -66.7 42 127.34 115.21 -52.28 diagonal 11. Corners (5p) with diverse 45 125.33 111.97 -67.12 40 122.37 110.41 -48.69 diagonal Note: g.1. = degree of freedom; G2 = Deviance; X2 = Pearson’s chi square; BIC = Bayesian Index Criteria. Source: Own elaboration from ENCEL 2017 (n=3,084). Nonetheless, model 3, which specifies different inheritance parameters according to occupational stratum, provides some distinct elements for both groups. The youths of the subgroup with higher treatment intensity have more favorable inheritance patterns than those belonging to the subgroup with lower intensity: if their origin is NM they are 4.07 times more likely to inherit their provider’s stratum than to be in another stratum (dp1=4.07, p<0.001), and if their origin is agricultural, they are 2.37 times more likely to inherit rather than have a destination in another stratum (dp6=2.37, p<0.001). This contrasts with the subgroup with lower treatment intensity, for which no evidence of intergenerational inheritance was found when their origin is NM (dp1=1.91, ls) but, in the case of origin being at the base of the stratification, the probability to inherit the provider’s stratum is 2.95 times higher than it is to be in another stratum (dp6=2.95, p<0.001), slightly higher than that of the high treatment intensity group. Although this evidence is limited and must be regarded with some caution (since the level of is so close to 1), the results suggest that there might be differences in favor of the group with high treatment intensity. Both groups are similar in terms of experiencing a strong intergenerational inheritance, but it seems that a higher intensity of PROSPERA’s treatment 40 favors a stronger inheritance at the top of the stratum (from where around 2% of the youths of the study group stem) and a weaker inheritance at the base (from where around 70% of the youths of the study group stem), in comparison to those who received a lower treatment intensity. If this is the case, a more intense treatment would contribute to an improved probability for youths with origins in those two strata of having a better socioeconomic status through their occupational position, although, this effect is considered modest regarding its contribution to achieving the interruption of intergenerational transmission of poverty. VI. Results of occupational attainment models This section presents the results of the structural models that analyzed the determinants of youths’ occupational attainment. First, we describe the general model, in which the study sample is analyzed, as a whole, and after that, the models that attempted to compare subgroups by sorting by sex, ethnic background and migratory status.54 A. General model Based on the model proposed in figure 1, we carried out tests to figure out if all our identified paths have empirical support. Four models, which are shown in Appendix H, were fitted and the resulting model includes those paths that were statistically significant. Figure 2 shows the results of the general model, with standardized coefficients, which allow for the comparison of the effect’s size between the different paths. The fit of the estimated model is considered satisfactory (RMSEA= 0.035; CFI=0.979; TLI=0.974; n=3,310).55 As it may be noted, the youth’s occupational attainment is explained by the indirect effects from their provider’s cognitive abilities, as well as the youth’s education and first occupation. In contrast with the initial model presented in figure 1, the selected model suggests that the provider’s cognitive abilities and social origin do not have direct effects on the current occupational attainment, but that its effects are expressed through mediating variables. Furthermore, the provider’s cognitive abilities have no direct influence on the youth’s education; and neither did social origin on the youth’s first occupation. Table 13 presents the standardized coefficients of the direct, indirect and total effects.                                                              54 We carried out the exercise of adjusting a model to compare the treatment intensity subgroups, however, the model did not converge when applying the weights described in section 4.3. 55 A structural model is considered to have good fit if the RMSEA is lower than 0.05 and the value of CFI and TLI is higher than 0.90. Furthermore, the adjusted model shows acceptable coefficients of determination for the endogenous latent variables (R2 between 0.30 and 0.56). Appendix D provides more details and references on the statistical fit, as well as the treatment of the missing values through FIML. 41 Figure 2. Standardized coefficients of the determinants of the occupational attainment model. Youths from 18-35 years of age of ENCEL 2017. Source: Own elaboration from ENCEL 2017. We can observe that the provider’s cognitive abilities and social origin have a positive and significant covariance ( =0.66, p<0.001), that is, a major/minor social origin is associated with better/worse cognitive abilities of the provider. As previously mentioned, both variables have indirect effects on the dependent variable, and are mediated by the other intervening variables. The provider’s cognitive abilities have a direct influence on the youth’s cognitive abilities, (1=0.40; p<0.001), which, in turn, have a direct effect on occupational attainment (6=0.08; p=0.001). On the other hand, social origin has a direct effect on the youth’s education (3=0.55; p<0.001) and this has a direct effect in the youth’s cognitive abilities (2=0.50; p<0.001), on the first occupation (5=0.70; p<0.001) and on the model’s dependent variable (7=0.16; p<0.001). Likewise, the youth’s first occupation has a direct effect on occupational attainment (8=0.32; p<0.001). A comparison of the total effects, that includes both the direct and indirect effects, allows us to understand the extent of the influence of each variable on occupational attainment. Regarding the variables associated with the ascribed factors, it is observed that the effect of social origin on occupational attainment is estimated to be 0.23, while those of the provider’s cognitive abilities are of 0.03. In total, both have an added effect of 0.26. The youth’s cognitive abilities have a total effect of 0.08 on the result variable, 0.42 on education, and 0.32 on first occupation, which equals an effect of 0.76 of the no-ascribed variables. 42 Table 13. Standardized effects of the determinants of occupational attainment. Youths from 18-35 years of age of ENCEL 2017. VARIABLES EFFECTS Social origin on occupational attainment Direct effect - Indirect effects 0.23 *** Via education 0.09 *** Via youth’s cognitive abilities and education 0.02 ** Via first employment and education 0.12 *** Total effect (direct + indirect) 0.23 *** Provider’s cognitive abilities on occupational attainment Direct effect - Indirect effects 0.03 ** Via youth’s cognitive abilities 0.03 Total effect (direct + indirect) 0.03 ** Youth’s cognitive abilities on occupational attainment Direct effect 0.08 *** Indirect effects - Total effect (direct + indirect) 0.08 *** Education on occupational attainment Direct effect 0.16 *** Indirect effects 0.26 *** Via youth’s cognitive abilities 0.04 ** Via first employment 0.22 *** Total effect (direct + indirect) 0.42 *** First occupation to occupational attainment Direct effect 0.32 *** Indirect effects - Total effect (direct + indirect) 0.32 *** Levels of significance *** p<0.001, ** p<0.01, * p<0.05. Source: Own elaboration from ENCEL 2017 (n=3,310). These findings lead us to conclude that social origin and the provider’s cognitive abilities are important factors that influence, through mediating factors, the youth’s occupational attainment. Nonetheless, the non-ascribed factors, on a whole, have an effect that is almost three times the effect of the ascribed-factors. Among the non-ascribed factors, the youth’s education is the factor that determines, with the greatest strength, the occupational attainment, followed by the first occupation and cognitive abilities. It is important to remember, however, that these three factors have been shaped by the two ascribed variables. Thus, both the ascribed as the non-ascribed factors intervene in the process, with the latter having a greater effect, which is consistent with the previous studies that indicate the importance of the origin-destination association, but there is also the possibility of other factors, mainly 43 education, to drive the processes of greater equality in opportunity. It is important to remember that these results not only suggest the importance of achieving a higher educational attainment, but also the importance of receiving schooling of better quality in order to improve occupational attainment. Furthermore, the influence of youths’ first occupation on the current occupational attainment highlights the need to delay the age of entering their first job and make sure that this employment has the highest possible status. B. The model by sex, ethnic background and migratory status The results for the analysis of the subgroups by sex, ethnic background and migratory are presented in this section. Figure 3 presents the standardized coefficients of the paths by sex, and the same notation is followed in figures 4 and 5 for the groups by ethnic background and migratory status, respectively. Tables 14 to 16 present the direct, indirect and total effects by sex, ethnic background and migratory status, respectively. The fits of the three estimated models are considered satisfactory, as can be observed from their fit measurements, shown in each figure. When the structural model is fitted according to sex, differences between females and males are observed in four paths (1, 3, 5, 8) (see figure 3). These results show us that females, in comparison with males, experience slightly weaker direct effects of the provider’s cognitive abilities on the youth’s abilities and slightly greater direct effects of social origin on education. At the same time, in comparison with males, females have a substantially weaker effect of education on their first occupation and experience an effect from this last variable on occupational attainment, while for males such effect is non-existent. The differences between these four specific paths are expressed in the comparison between the indirect and total effects by sex (see table 14), according to which presents differences in the three variables. First, it shows that females, in comparison with males, experience total effects from social origin on occupational attainment that are 30% higher. These differences are explained by the strength of the routes they pursue through education, as well as through education and first occupation. Second, although the total effect of education on occupational attainment is statistically similar, in the case of females, we observe a greater direct influence from education on occupational attainment and indirect influence through their first employment. Finally, as previously mentioned, first employment influences current occupational attainment more. 44 Figure 3. Standardized coefficients of the determinants of occupational attainment model. Youths from 18-35 years of age of ENCEL 2017. Comparison by sex. Note: The estimated coefficients are denoted for each path as i(female/male). The dotted line denotes that the difference between the estimated coefficients when comparing them between the groups is non-significant, while the solid line indicates that the difference between groups is significant. For example, 2=.49***/.50*** means that the direct effect of education on the youth’s cognitive abilities is significant for males and females and there is no difference between groups. Source: Own elaboration from ENCEL 2017. Thus, although social origin has more weight on females, education plays a more important role for them, which compensates for the effects of social origin and may mitigate the strength of inheritance, in the event they achieve higher schooling and attend schools with better quality. This suggests they experience greater equality of opportunity and, possibly, also a greater ability to compensate for the higher weight of their social origin. Nonetheless, the importance of first occupation indicates that females face greater difficulty in order to improve their occupational status throughout their life trajectory, even though, as previously mentioned, they have a higher status. 45 Table 14. Standardized effects of the determinants of occupational attainment. Youths from 18-35 years of age of ENCEL 2017. Comparison by sex. Females Males Groups Variables Effects Effects Comparison Social origin on occupational attainment Direct effect - - Indirect effects 0.27 *** 0.19 *** *** Via education 0.12 ** -0.52 ls * Via education and cognitive abilities 0.03 * 0.03 * ns Via education and first employment 0.13 *** 0.68 ls * Total effect (direct + indirect) 0.27 *** 0.19 *** *** Provider’s cognitive abilities on occupational attainment Direct effect - - Indirect effects 0.04 * 0.04 * ns Via youth’s cognitive abilities 0.04 * 0.04 * ns Total effect (direct + indirect) 0.04 * 0.04 * ns Youth’s cognitive abilities on occupational attainment Direct effect 0.1 * 0.1 * ns Indirect effects - - Total effect (direct + indirect) 0.1 * 0.1 * ns Education on occupational attainment Direct effect 0.21 ** -0.96 ls * Indirect effects 0.27 *** 1.31 ls *** Via youth’s cognitive abilities 0.05 * 0.05 * ns Via first employment 0.23 *** 1.26 ls * Total effect (direct + indirect) 0.48 *** 0.35 *** ns First occupation to occupational attainment Direct effect 0.34 * 1.34 ls * Indirect effects - - Total effect (direct + indirect) 0.34 * 1.34 ls * Levels of significance *** p<0.001, ** p<0.01, * p<0.05. Source: Own elaboration from ENCEL 2017 (n=3,310). Regarding the ethnic origin, differences between indigenous and non-indigenous in three paths (1, 2, and 3) were observed (see figure 4). In the indigenous group, compared with the non-indigenous, the provider’s cognitive abilities have a weaker effect on the youth’s cognitive abilities, social origin influences the youth’s education with slightly more strength and the latter with greater magnitude on the youth’s cognitive abilities. Overall, this appears to indicate that indigenous youths experience a slightly less influence from the ascribed factors on their cognitive abilities and more on their education. 46 Figure 4. Standardized coefficients of the determinants of occupational attainment model. Youths from 18-35 of age of ENCEL 2017. Comparison by ethnic background. Note: The estimated coefficients are denoted for each path as (indigenous/non-indigenous). The dotted line denotes that the difference between the estimated coefficients, when comparing them between the groups, is non-significant, while the solid line indicates that the difference between groups is significant. For example, 2=.63***/.49*** means that the direct effect of education in the youth’s cognitive abilities is significant for indigenous and non-indigenous youths and there is a significant difference between groups. Source: Own elaboration from ENCEL 2017. Nonetheless, when comparing indirect and total effects of each variable on occupational attainment (see table 15), we observe that there are practically no differences between the groups, except for the indirect effect of the provider’s cognitive abilities on occupational attainment via the youth’s cognitive abilities. However, although the difference is statistically significant, the magnitude is very close to zero. Therefore, in summary, we observe that for both groups there is an effect of similar magnitude of all variables on the youth’s occupational attainment. Thus, the aforementioned differences between the groups on the determination of the cognitive abilities, are not translated into differentiated processes of occupational attainment for the indigenous and non-indigenous youths. 47 Table 15. Standardized effects of the determinants of occupational attainment. Youths from 18-35 years of age of ENCEL 2017. Comparison by ethnic background. Indigenous Non-indigenous Groups Variables Effects Effects Comparison Social origin on occupational attainment Direct effects - - Indirect effects 0.25 *** 0.24 *** ns Via education 0.10 * 0.10 ** ns Via education and cognitive 0.02 ls 0.02 * ns abilities Via education and first 0.13 *** 0.12 *** ns employment Total effect (direct + indirect) 0.25 *** 0.24 *** ns Provider’s cognitive abilities on occupational attainment Direct effect - - Indirect effects 0.01 ls 0.03 * * Via youth’s cognitive abilities 0.01 ls 0.03 * * Total effect (direct + indirect) 0.01 ls 0.03 * * Youth’s cognitive abilities on occupational attainment Direct effect 0.06 ls 0.08 * ns Indirect effects - - Total effect (direct + indirect) 0.06 ls 0.08 * ns Education on occupational attainment Direct effect 0.17 * 0.18 ** ns Indirect effects 0.25 *** 0.25 *** ns Via youth’s cognitive abilities 0.04 ls 0.04 * ns Via first employment 0.21 *** 0.22 *** ns Total effect (direct + indirect) 0.42 *** 0.43 *** ns First occupation to occupational attainment Direct effect 0.31 *** 0.31 *** ns Indirect effects - - Total effect (direct + indirect) 0.31 *** 0.31 *** ns Levels of significance *** p<0.001, ** p<0.01, * p<0.05. Source: Own elaboration from ENCEL 2017 (n=3,310). The comparison of migratory statuses shows differences in five paths (1, 2, 5, 6 and 7) (see figure 5). The direct effects of the youth’s education on the cognitive abilities and occupational attainment are greater for migrants than for those who remained in the localities of origin. On the contrary, for the non-migrants group, the direct effects of the provider’s cognitive abilities are greater on the youth’s cognitive abilities, and of these on occupational attainment and of education on youth’s first employment. 48 Figure 5. Standardized coefficients of the determinants of occupational attainment model. Youths from 18-35 years of age of ENCEL 2017. Comparison by migratory status. Note: The estimated coefficients are denoted for each path as (migrant/non-migrant). The dotted line denotes that the difference between the estimated coefficients, when comparing them between the groups, is non-significant, while the solid line indicates that the difference between groups is significant. For example, 2=.59***/.44*** means that the direct effect of education in the youth’s cognitive abilities is significant for migrants and non-migrants and there is a significant difference between the groups. Source: Own elaboration from ENCEL 2017. When reviewing the indirect and total effects (see table 16), social origin was shown to have stronger effects on occupational attainment in the case of migrants, while provider’s cognitive abilities had stronger effects on non-migrants. If the effects of both ascribed variables for each subgroup are added, the effects of the ascribed variables on occupational attainment for migrants and non-migrants are similar. In relation to the non-ascribed effects, for migrants, their cognitive abilities have no influence on occupational attainment, but their education has a 21% greater effect on occupational attainment than non-migrants. In summary, differentiated effects of migratory status were observed. The effects of the ascribed variables on occupational attainment have similar strength, although, in the case of migrants, the variable with more weight was social origin and for non-migrants it was their provider’s cognitive abilities. Perhaps, the most important difference is that those who migrated from their localities of origin presented considerably higher effects from education on their current occupational attainment. This may lead us to think that they experience a greater equality of opportunity, through having more possibility to modify the effect of their social origins on 49 their occupational attainment than non-migrants by way of taking advantage of the education received in contexts different from those of their place of origin. Table 16. Standardized effects of the determinants of occupational attainment. Youths from 18-35 years of age of ENCEL 2017. Comparison by migratory status. No Migrants Groups Variables migrants Effects Comparison Effects Social origin on occupational attainment Direct effects - - Indirect effects 0.27 *** 0.21 *** *** Via education 0.13 * 0.06 * ** Via education and cognitive abilities 0.01 ls 0.02 ** ns Via education and first employment 0.13 *** 0.13 *** ns Total effect (direct + indirect) 0.27 *** 0.21 *** *** Provider’s cognitive abilities on occupational attainment Direct effect - - Indirect effects 0.01 ls 0.05 * ** Via youth’s cognitive abilities 0.01 ls 0.05 * ** Total effect (direct + indirect) 0.01 ls 0.05 * ** Youth’s cognitive abilities on occupational attainment Direct effect 0.04 ls 0.10 ** * Indirect effects - - Total effect (direct + indirect) 0.04 ls 0.10 ** * Education on occupational attainment Direct effect 0.23 * 0.11 * ** Indirect effects 0.25 *** 0.27 *** ns Via youth’s cognitive abilities 0.02 ls 0.04 ** ns Via first employment 0.23 *** 0.22 *** ns Total effect (direct + indirect) 0.48 *** 0.38 *** *** First occupation to occupational attainment Direct effect 0.35 *** 0.31 *** ns Indirect effects - - Total effect (direct + indirect) 0.35 *** 0.31 *** ns Levels of significance *** p<0.001, ** p<0.01, * p<0.05. Source: Own elaboration from ENCEL 2017 (n=3,310). VII. Conclusions Throughout its history, PROSPERA has been consolidated as one of the central governmental interventions in the country directed at improving the quality of life of the population living in conditions of poverty. The program’s design has focused on two objectives, improving the current 50 consumption of the families living in conditions of extreme poverty and developing the human capital of their members, specifically children and youths. All this with the aim, through the insertion of youths into the labor market under better conditions, of contributing towards the breaking of the intergenerational inheritance of poverty. In this sense, the program’s actions aim to mitigate the effects of adverse social origins on the fate of children and youths, by developing their human capital and entrance into the labor market, in order to achieve social mobility. PROSPERA has proven to have positive impacts in the short and medium-term on the different areas of welfare of the beneficiary families. These effects reflect an improvement in various indicators, such as current consumption, nutrition, health and education, among others. Nonetheless, limitations regarding the development of human capital are also emphasized. For example, the quality of the schools to which beneficiary children and youths have access is deficient and, in spite of the positive impacts on education, youth beneficiaries of the program still present significant gaps in the number of years of schooling and the level of learning achieved compared to their counterparts at the national level. The evidence, to date, is less extensive in terms of breaking the intergenerational inheritance of poverty. The studies performed so far, have mostly focused on the effects from the program after 10 years of intervention in rural areas. These studies have approached the subject through the analysis of the characteristics of the youth’s occupational mobility, since not enough time has elapsed to directly analyze whether the inheritance of poverty between generations has been interrupted (or not). The results of the research on labor experience of the rural beneficiary youths of the program have been mixed. It has been shown that the characteristics of their labor are more precarious than those of youths at the national level and that, although registering positive impacts in some labor indicators (labor participation, hours worked, labor income and formal employment), their magnitudes are considered modest. Additionally, no impact of PROSPERA on intergenerational occupational mobility was identified. The implementation of ENCEL 2017 opens up the possibility to generate new knowledge on the situation of the youth beneficiaries of PROSPERA, two decades after the intervention started, in order to estimate the program’s effectiveness at achieving its ultimate goal. This study is entering into this juncture and aims to contribute to the same thorough analysis of two strands: intergenerational occupational mobility and the occupational attainment process of the rural beneficiary youths of PROSPERA. For both, a group of youth beneficiaries of the program, between 18 and 35 years of age, and who are employed in the labor market, was studied. General characteristics were analyzed, as well as differences between the subgroups distinguished by sex, ethnic background and migratory status. Furthermore, we analyzed whether a higher intensity of PROSPERA’s treatment received, associated with the size and structure of the household, increased quality in labor opportunities for youths, expressed both in higher upward mobility rates, as well as in a greater social fluidity. The estimation carried out compares two groups with a 7% difference in time of exposure to the program and 34% difference in the volume of cash transfers received by the households. 51 An initial comparison of some educational and labor characteristics of our sample of youths with youths in the same age range both at a national level and in only rural areas, shows that this group of PROSPERA beneficiaries have disadvantageous conditions: they had lower levels of school attendance, number of years of schooling, labor income, occupational status and access to health services. When contrasting the characteristics of the different subgroups of interest, it showed that women have more schooling and occupations with higher status, but lower labor income and less access to health services, which is consistent with the gender inequalities that have been extensively documented before. It should be emphasized that indigenous people, in comparison to non-indigenous, have lower labor remunerations and less access to health services, which may be associated not only with ethnic discrimination, but also to the fact that their residences are, in greater proportion, in localities with very high marginalization. Migrants, in contrast with non-migrants, have more schooling, occupations with higher status, a higher labor income and greater access to health services. This may be associated with the change in context of labor insertion, since youths who migrate from their place of origin now reside in predominantly urban localities (half of them are in metropolitan areas), with lower levels of marginalization than their localities of origin. An analysis of intergenerational mobility emphasizes the strength and pattern of association between the origins and destinations of individuals. The people in our study’s sample, as beneficiaries of PROSPERA, come from disadvantaged social origins. In the area of occupation, this is reflected in a higher proportion of youths (7 out of 10 cases) with an occupational origin within the agricultural stratum, which is the bottom step of our proposed stratification. A movement towards occupational strata of higher hierarchy entails the access by individuals to better life opportunities. From their position of origin, half of the studied youths experienced absolute upward mobility, while one-quarter inherited their provider’s occupation and the rest moved downward to a stratum of lesser hierarchy. Although upward mobility rates have a more significant magnitude, we also saw that the probability of having a destination at the bottom of the stratification is very high for youths of all origin strata, while the probability of having a destination at the top was low, except for those whose origin are already located there. The relative mobility analysis provided additional evidence in this regard, since it showed that the association between the youths’ occupational origin and destination, once the structural changes are controlled for, is mainly determined by intergenerational inheritance, which is stronger at both extremes of the occupational stratification. That is, those who stem from the highest and lowest strata have a high probability of remaining where they begin. The barrier in long-distance occupational upward mobility for those who have their origin in the agricultural stratum is added to this, which increases the difficulty for upward social mobility of youths with more disadvantaged origins. When observing the behavior of the absolute mobility among the analyzed subgroups, the differences by sex and, even more, by migratory status are emphasized. Women and migrants experience higher upward mobility rates, in contrast to a greater immobility by men and non- migrants, respectively. This is also reflected on the fact that women and migrants are less likely 52 to have an occupation in the agricultural stratum and are more likely to be in a NM occupation (for any stratum of origin), in comparison to men and non-migrants, respectively. The results of the relative mobility analysis confirm that both women, in comparison to men, and migrants, in contrast to non-migrants, have more social fluidity or equality in labor opportunities. Women experience less strength from occupational inheritance, and they do not face a barrier in long- distance upward mobility from the base of the stratification. Migrants, for their part, do not experience occupational inheritance, nor barriers in long-distance upward mobility from the agricultural stratum. A comparison of ethnic backgrounds shows a small difference in absolute mobility rates. There is only a slightly higher probability of indigenous youths of having an agricultural destination when their provider had an occupation in the four strata with a lesser position in the hierarchy is observed, while they are more likely than those not-indigenous of inheriting their provider’s occupation if their origin is NM. Paradoxically, this would suggest that, although it is more likely that indigenous people, in comparison to non-indigenous, occupy the stratum of lower hierarchy, it is also more likely that they inherit their provider’s occupation when they come from a higher stratum. Nonetheless, the relative mobility analyses did not provide evidence that suggests differences in the social fluidity or equality in labor opportunities between indigenous and non- indigenous youths. Regarding the comparisons of different treatment intensity, no substantial differences in the absolute mobility rates are observed between both groups, which may be partially due to the relative similarity in intensity with which our comparison groups participated of the program. In terms of relative mobility, it is important to note that the two groups experienced intergenerational inheritance, which has a differentiated strength for the different strata of origin, being stronger at both extremes. Nonetheless, youths of the high intensity group experience a greater probability of intergenerational inheritance when their origin is in the MN stratum and less probability of inheritance when they stem from the agricultural stratum. Therefore, although the findings must be considered with caution, they suggest that a higher treatment intensity of PROSPERA may raise the chances of youths with origins in the NM stratum to remain at the top of this hierarchy and would help to facilitate upward occupational mobility from the base of the occupational hierarchy. If this is the case, a more intense treatment would contribute to raising the likeliness of youths to have occupations related to better socioeconomic levels, although this effect is considered modest regarding its contribution towards breaking the intergenerational transmission of poverty. In addition, the study of the determinants of occupational attainment contributes to the knowledge of the micro-social processes that lead to the macro-social level results reflected in the absolute mobility rates, as well as the features of the relative mobility. Specifically, it assists in the identification of the direction and strength of the relationship between some of the factors that influence youths’ occupational status. From a concern for equality in opportunity, the usual interest is to discern the magnitude of the effects of the factors that are associated to social origins 53 or considered ascribed, in contrast to those that are not and that may mediate the relationship between social origins and occupational destinations. Our findings, in this regard, complement this description. The social origin of youths and their provider's cognitive abilities, considered ascribed variables, have an important impact on the youths’ occupational destination through their effect on youths’ cognitive abilities, education and first occupation. Nonetheless, these non-ascribed variables, as a whole, affect the status of the youths’ current occupation even more than the ascribed variables. Among non-ascribed factors, the youth’s education is the determinant factor of greatest strength, followed by the youths’ first occupation and their cognitive abilities. In summary, the disadvantageous social origins of these youths continue to represent an obstacle to them in achieving their best labor insertion, which may help explain the high rates of occupational inheritance. However, more and better education, entering first employment at an older age, a better first occupation and greater cognitive abilities of youths, may have important effects towards improving their occupational destination and promoting upward mobility, as well as equality in labor opportunities. The study of occupational attainment through subgroups allows us to emphasize some relevant differences. In comparisons with men, women experienced stronger effects of social origin on their occupational attainment, which are compensated by also experiencing greater effects from education. This contributes towards understanding the higher rates of upward mobility and greater social fluidity of women, which, as a whole, suggests a greater equality in labor opportunities among studied females. Nonetheless, gender segregation in the labor market is also at stake, encouraging women to leave the agricultural occupation at a greater proportion than men. Furthermore, as mentioned above, a greater labor status for women does not necessarily correspond to greater labor income, given the gender inequality in labor remunerations. A greater occupational inheritance of men may be related to the higher effect of their first occupation on the occupational destination, which would hint towards a lower mobility within their own labor trajectory. In relation to the subgroups of different ethnic backgrounds, the similarities in absolute mobility and the evidence that points towards an absence of differences in relative mobility are consistent with the similarity of the process of occupational attainment in both subgroups. It is remarkable that no important differences in these aspects were found between both subgroups. It could be thought that, when analyzing the population living in conditions of extreme poverty, that ethnic differentiation plays a less important role than when in comparison to at a national level. On the other hand, this absence of differences may be considered as a favorable result, taking into account that, in comparison with non-indigenous, a greater proportion of indigenous population originates from and currently have their residence in localities with more adverse conditions. The differences in the process of occupational attainment between youths that migrated from their localities of origin and those who reside in the same locality of their origin helps to explain the gap in their mobility rates and the higher social fluidity of migrants. It was observed that the 54 effects of the ascribed variables on the youths’ occupational destination are of similar strength. Nonetheless, the most important difference is that, those who migrate from their localities of origin demonstrate considerably higher overall effects from their education on their occupational attainment, which refers to a greater equality in opportunity. It is likely that moving to mainly urban localities, with a lower level of marginalization and with likely more dynamic labor markets, allows migrants to effectively apply the education they have received and, thus, obtain occupations of higher status, which represents upward mobility and access to better life opportunities. The described results show an overview of the contrasts present in the youths’ group studied. Half were able to ascend with respect to their occupation of origin (and, therefore, the other half did not), but even then, their occupational insertion is not favorable, they have strong occupational inheritance and barriers to social upward mobility. This experience is not homogeneous; women and migrants presented greater equality in their labor opportunities. The evidence points to the importance of education, first occupation and the youths’ cognitive abilities (in that order) as mediating factors that may mitigate the weight of social origins on their occupational destinations. Furthermore, the findings suggest that the characteristics of the contexts of labor insertion are important. In this sense, public policy should reinforce these areas. As indicated in previous studies, it is necessary to encourage the cognitive development of children, extend their schooling trajectories, improve the quality of the educational institutions they attend and promote greater learning. Also, strategies to delay the age of youths’ first employment and improve the status of their first occupation are required, as well as to link their spheres of education and labor. Additionally, it is essential to promote the development of rural localities, as well as the regions they belong to. These findings described are similar to those reported in studies on intergenerational occupational mobility performed after 10 years of intervention of the program in rural areas (see Yaschine, 2015). The main difference is that, while in the medium-term, no effect from the program on the inequality of labor opportunities was identified, in this study the findings suggest a possible effect from receiving a higher intensity of PROSPERA’s treatment. Nonetheless, the samples from both studies are not strictly comparable and the current results are not conclusive, consequently, they must be taken with a grain of salt. The comparison we carried out suggests that a higher treatment intensity may generate a greater probability for youths to have occupations related to better socioeconomic levels, through a greater inheritance of NM positions and a lesser inheritance of the agricultural stratum. However, in the backdrop of this distinction are the high intergenerational inheritance rates for both groups, which describes a less encouraging picture. Due to the limitations of the sample, we consider that the results of this study are valid only for the subpopulation analyzed, although they may be considered as indicative for the beneficiary population of PROSPERA in rural areas. Furthermore, the estimated effects of PROSPERA reflect a reduced difference in treatment intensity and should not be taken as the overall effect that the program may have after 20 years. Additionally, the estimated effects may be underestimated for two reasons. First, ENCEL 2017 did not collect information on all migrants 55 that comprised the selected sample and, as we have seen, those who migrated from their localities of origin are those who achieved more favorable labor results. Second, our study group consisted of a majority that still have yet to reach labor maturity; which makes us think that they may have a better occupational status in the future. This is only a hypothetical assessment since, as mentioned above, the analyses we performed by age groups did not show better labor results for older individuals. It is important to remember at this point that the processes of occupational attainment studied existed in an unfavorable economic, labor and social context, which constituted a weak and unequal structure of opportunities. This, without a doubt, has limited the access of the study group’s youths to different types of economic and social assets and services, as well as to labor opportunities, specifically, in the rural localities where they largely stem from. This scenario has clearly not been favorable for the occupational mobility of the beneficiaries of PROSPERA or the leveling opportunities that could promote the process of breaking the intergenerational inheritance of poverty. The challenges that arise from here involve the program, but also extend beyond it. The intergenerational cycle of poverty may only be interrupted by the implementation of a multiplicity of public policy actions. As PROGRESA anticipated during its introduction, to achieve this purpose depends, among other factors, on proper macroeconomic performance and a dynamic labor market, which have not yet materialized. Furthermore, the international experience suggests that a welfare system with strong, universal features and policies that equalize opportunities is required, specifically in the childhood and youth. These challenges call for a comprehensive public policy, of which PROSPERA would be one of the pieces, that has the reduction of inequality and poverty as its objectives.   56 References   Acevedo, Ivonne; Ortega, Araceli and Székely, Miguel. "Rendimiento Escolar y Transiciones Laborales con Transferencias Condicionadas en México." Manuscript (2018). Agresti, Alan. An Introduction to Categorical Data Analysis. Hoboken, New Jersey: John Wiley & Sons, Inc., 2017 Agudo, Alejandro. "¿Cómo se Explica el Impacto Educativo del Programa Oportunidades? Actores, Factores y Procesos." In A Diez Años de Intervención. Evaluación Externa del Programa Oportunidades 2008 en Zonas Rurales (1997-2007), 71-169. Mexico, D.F.: INSP, 2008. Alderton, David L., & Larson, Geralt E. "Dimensionality of Raven´S Advanced Progressive Matrice Items." Educational and Psychological Measurement, no. 50 (1990): 887-900. Arbuckle, James L. "Full Information Estimation in the Presence of Incomplete Data." In Advanced Structural Equation Modeling: Issues and Techniques, G.A. Marcoulides & R.E. Schumacker, 243-77. Mahwah, New Jersey, 1996. Attanasio, Orazio; Meghir, Costats and Schady, Norbert. "Mexico's Conditional Cash Transfer Programme." The Lancet 375, no. 9719 (2010): 980. Beller, Emily and Hout, Michael. "Welfare States and Social Mobility: How Educational and Social Policy May Affect Cross-National Differences in the Association between Occupational Origins and Destinations." Research in Social Stratification and Mobility 24 (2006): 353-65. Bentler, Peter M. "Comparative Fit Indexes in Structural Models." Psychological Bulletin, no. 107 (1990): 238-46. Bilker, Warren B; Hansen, John A; Brensinger, Colleen M; Richard, Jan; Gur, R.E & Gur, Raquel C. "Development of Abbreviated Nine-Item Formas of the Raven´S Standard Progressive Matrices Test." Assesment 19, no. 3 (2012): 354-69. Bird, Kate. The Intergenerational Transmission of Poverty. An Overview. Work document 99, Manchester: Chronic Poverty Research Center, 2007. Blanco, Emilio. Los Límites de la Escuela. Educación, Desigualdad y Aprendizajes en México. Mexico: El Colegio de México, 2011. Blau, Peter y Duncan, Otis. "The Process of Stratification ". In Social Stratification in Sociological Perspective. Class, Race and Gender, edited by D Grusky, 390-403. Colorado: Westview Press, 2001. Box, George Edward P. "Science and Statics." Journal of the American Statistical Association, no. 71 (1976): 791-99. Bradley, Robert H. & Corwyn, Robert F. "Caring for Children around the World: A View from Home." International Journal of Behavioral Development 29, no. 6 (2005): 468-78. Breen, Richard. "The Comparative Study of Social Mobility." In Social Mobility in Europe, R. Breen, 1- 16. Oxford: Oxford University Press, 2004a. ———. "Foundations of Neo-Weberian Class Analysis." In Approaches to Class Analysis, E.O. Wright, 31-50. Cambridge: Cambridge Univestity Press, 2005. ———. "Statistical Methods of Mobility Research." In Social Mobility in Europe, R Breen, 17-35. Oxford: Oxford University Press, 2004b. Breen, Richard and Jonsson, Jan. "Inequality of Opportunity in Comparative Perspective: Recent Research on Educational Attainment and Social Mobility." Annu, Rev. Sociol 31 (2005): 223-43. Breen, Richard and Luijkx, Ruud. "Conclusions." In Social Mobility in Europe, R. Breen, 384-410. Oxford: Oxford University Press, 2004. Brown, Timothy. Confirmatory Factor Analysis for Applied Research. 2nd ed. New York: The Guilford Press, 2015. 57 Browne, Michael W. & Cudeck, Robert. "Alternative Ways of Assessing Model Fit ". In Testing Structural Equation Models, K. A. Bollen, & Long J. S., 136-62. Newbury Park, CA: Sage, 1993. Campos, Pilar. "Documento Compilatorio de la Evaluación Externa 2008 del Programa Oportunidades." Mexico: Oportunidades, 2010. CEEY. "Informe Movilidad Social en México 2013. Imagina tu Futuro.". Mexico, D.F.: CEEY, 2013. Consejo Nacional de Evaluación de la Política de Desarrollo Social (CONEVAL). (n/d)."Evolución de la Pobreza por la Dimensión de Ingreso, 1992-2016." ———"Informe de Evaluación de la Política de Desarrollo Social 2018." CONEVAL. Mexico, 2018. Cortés, Fernando. "La Relación entre el Gasto Social y la Pobreza a Debate." In Temas de Política Social de México y América Latina, F. Cortés, 117-57. Mexico: El Colegio de México y UNAM, 2018. Cortés, Fernando and Escobar, Agustín. "Modelos de Acumulación de Capital y Movilidad Social. Un Estudio en Seis Ciudades Mexicanas." In Cambio Estructural y Movilidad Social en México, F. Cortés; Escobar, A. y Solís, P., 21-74. Mexico: El Colegio de México, 2007. Cortés, Fernando and Rubalcava, Rosa María. "El Progresa como Respuesta a la Crisis De 1994 ". In Transferencias Condicionadas y Sociedad, M. González de la Rocha and A. Escobar, Mexico D.F: CIESAS, 2012. Cortés, Fernando and Vargas, Delfino. Origen Es Destino: Un Análisis Longitudinal de la Marginación Municipal. México 1990-2015 Mexico D.F.: Siglo XXI editores-UNAM, 2017. Cortés, Fernando; Ochoa, Sara; Vargas, Delfino and Yaschine, Iliana. "La Desigualdad en la Distribución del Ingreso en los ODS México a 2030.". Perspectivas del Desarrollo a 2030, R. Cordera, and E. Provencio, 269-85. Mexico: UNAM-PUED, 2016. CPRC. "The Chronic Poverty Report 2004-2005." Manchester: Institute for Development Policy and Management-Chronic Poverty Research Centre, 2004. Cruz, Carlos; De la Torre, Rodolfo and Velásquez, César. "Evaluación Externa de Impacto del Programa Oportunidades 2001-2006. Informe Compilatorio." Mexico: Instituto Nacional de Salud Pública, 2006. CTPM. "Medición De La Pobreza: Variantes Metodológicas y Estimación Preliminar." Ministry of Social Development. México, 2002. Chou, Chih-Ping & Bentler, Peter M. " Estimates and Tests in Structural Equation Modeling ". In Structural Equation Modeling: Concepts Issues and Applications, edited by R. H. Hoyle, 37-55. Thousand Oaks, CA: Sage, 1995. Dubet, Francoise. Repensar la Justicia Social: Contra el Mito de la Igualdad de Oportunidades. Buenos Aires: Siglo XXI editores, 2014. Erikson, Robert and Goldthorpe, John. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press, 1992. ———. "Trends in Class Mobility: The Postwar European Experience." In Social Stratification in Sociological Perspective. Class, Race and Sex, D. Grusky, 334-72. Boulder: Westview Press, 2001. Evans, Gary W. y Schamberg, Michelle A. "Childhood poverty, chronic stress and adult working memory." Proceedings of the National Academy of Sciences of the United States of America 106, 16 (2009): 6545-6549. Faulkner, William N. "A Critical Analysis of a Randomized Controlled Trial Evaluation in Mexico: Norm, Mistake or Exemplar?". Evaluation 20, no. 2 (2014): 230-43. Fernald, Lia CH; Gertler, Paul J and Neufeld, Lynnette M. "10-Year Effect of Oportunidades, Mexico's Conditional Cash Transfer Programme, on Child Growth, Cognition, Language, and Behaviour: A Longitudinal Follow-up Study." The Lancet 374, no. 9706 (2009): 1997-2005. ———. "Mexico's Conditional Cash Transfer Programme–Authors' Reply." The Lancet 375, no. 9719 (2010): 981. 58 Fong, Christian; Ratkovic, Marc & Imai, Kosuke. "Cbps: R Package for Covariate Balancing Propensity Score." Comprehensive R Archive Network (CRAN) (2014). Frangakis, Constantine E.; Rubin, Donald B. and Zhou, Xiao‐Hua. "Clustered Encouragement Designs with Individual Noncompliance: Bayesian Inference with Randomization, and Application to Advance Directive Forms." Biostatistics 3, no. 2 (2002): 147-64. Ganzeboom, Harry and Treiman, Donald. "International Stratification and Mobility File: Conversion Tools." Department of Social Research Methodology, http://www.harryganzeboom.nl/ismf/index.htm (last revision January 6, 2012). Consulted on September 24, 2018. ———. "Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Ocupations." Social Science Research 25 (1996): 201-39. Ganzeboom, Harry; De Graaf, Paul and Treiman, Donald. "A Standard International Socio-Economic Index of Occupational Status." Social Science Research 21 (1992): 1-56. Ganzeboom, Harry; Treiman, Donald and Ultee, Wout. "Comparative Intergenerational Stratification Research: Three Generations and Beyond." Annu, Rev. Sociol 17 (1991): 201-39. Goldthorpe, John. "Class Analysis and the Reorentation of Class Theory: The Case of Persisting Differentials in Educational Attainment." The British Journal of Sociology 43, no. 3 (1996): 481-505. ———. "Progress in Sociology: The Case of Social Mobility Research." Analyzing Inequality. Life Chances and Social Mobility in Comparative Perspective, S. Svallfors, 56-82. Stanford: Stanford University Press, 2005. González de la Rocha, Mercedes and Escobar, Agustín. "La Vida Después de Oportunidades: Impacto del Programa a Diez Años de su Creación." In A Diez Años de Intervención. Evaluación Externa del Programa Oportunidades 2008 En Zonas Rurales (1997-2007), 125-98. Mexico, D.F.: INSP, 2008. Goodman, Leo A. "Exploratory Latent Structure Analysis Using Both Identifiable and Unidentifiable Models." Biometrika 61, no. 2 (1974): 215-31. Gough, Ian; McGregor, Allister and Camfield, Laura. Wellbeing in Developing Countries. Conceptual Foundations of the Wed Programme. Work document of Wed 19 Bath: University of Bath, 2006. Hackman, Daniel A., Gallop, Robert, Evans Gary W. y Farah Martha J. "Socioeconomic status and executive function: developmental trajectories and mediation." Developmental Science 18, 5 (2015): 686- 702. Hauser, Robert and Warren, John Robert. "Socioeconomic Indexes for Occupations: A Review, Update and Critique." In Social Stratification in Sociological Perspective. Class, Race and Gender, D. Grusky, 281-86. Boulder: Westview Press, 2001. Hernández, Daniel. Historia de Oportunidades. Inicios y Cambios del Programa. Mexico: FCE, 2008. Hout, Michael. Mobility Tables. Series on Quantitative Applications in the Social Sciences. Beverly Hills: Sage Publications, 1983. Hout, Michael and DiPrete, Thomas. "What We Have Learned: Rc28´S Contributions to Knowledge About Social Stratification." Research in social stratification and mobility 24 (2006): 1-20. Ibarrarán, Pablo and Villa, Juan Manuel. "Labor Insertion Assessment of Conditional Cash Transfer Programs: A Dose-Response Estimate for Mexico's Oportunidades." In XIV Meetings of the LACEA/ IADB/ WB/ UNDP Research Network on Inequality and Poverty (NIP). University of Tulane, New Orleans, 2010. IFPRI. "¿Está Dando Buenos Resultados Progresa? Informe de los resultados de una evaluación realizada por el IFPRI." Progresa. Mexico, 2000. Imai, Kosuke & Ratkovic, Marc. "Covariate Balancing Propensity Score." Journal of the Royal Statistical Society: Series B (Statistical Methodology) 76, no. 1 (2014): 243-63. INSP. "Nota Técnica sobre el Diseño muestral para la Encuesta de Evaluación Rural de Oportunidades (Encel 2007)." Manuscript. Mexico: Instituto Nacional de Salud Pública, 2007. 59 Jeffreys, Harold. Theory of Probability. 3rd ed. London: Oxford University Press, 1961. Kaztman, Rubén and Filgueira Carlos. Marco Conceptual sobre Activos, Vulnerabilidad y Estructura De Oportunidades. Montevideo: Comisión Económica para América Latina y el Caribe, 1999. Kerbo, Harold. Social Stratification and Inequality: Class Conflict in Historical, Comparative and Global Perspective. New York: McGraw-Hill, 2006. Kerckhoff, Alan. "Institutional Arrangements and Stratification Processes in Industrial Societies." Annu, Rev. Sociol 15 (1995): 323-47. Kugler, Adriana and Rojas, Ingrid. "Do Ccts Improve Employment and Earnings in the Very Long- Term? Evidence from Mexico." NBER Working Paper, no. 24248 (2018). Levy, Santiago and Rodríguez, Evelyne. "El Programa de Educación, Salud y Alimentación, Progresa ". In Ensayos sobre el Desarrollo Económico y Social de México, S. Levy, 181-379. Mexico: FCE, 2004. Lin, Nan. "Social Networks and Status Attainment." Annu, Rev. Sociol 25 (1999): 467-87. Luiselli, Cassio. Agricultura y Alimentación en México. Evolución, Desempeño y Perspectivas. Mexico: Siglo XXI Editores, 2017. Mancera, Carlos; Priede, Alejandra and Serna, Leslie. "El Desempeño de los Becarios del Programa Oportunidades en la Prueba Enlace: Cambios entre 2008 y 2011 en Educación Básica y Media Superior." Manuscript (2012). Mancera, Carlos; Serna, Leslie and Priede, Alejandra. "Modalidad Educativa y Organización Multigrado como Factores Asociados a Brechas de Aprendizaje de Becarios de Oportunidades, Primaria y Secundaria 2007." In A Diez Años De Intervención. Evaluación Externa Del Programa Oportunidades 2008 En Zonas Rurales (1997-2007), 15-70. Mexico, D.F.: INSP, 2008. Martínez, Jesuswaldo and Rosales, Maritza. "Decisiones de Política Pública para Elevar el Empleo. Informe del Desarrollo en México." In Propuestas Estratégicas para el Desarrollo 2019-2024, R. Cordera, and E. Provencio, Mexico: PUED-UNAM, 2018. McDonald, Roderick P. Test Theory: A Unified Treatment. Mahwah: Lawrence Erlbaum Associates, Inc., 1999. Moore, Karen. Frameworks for Understanding the Inter-Generational Transmition of Poverty and Well- Being in Developing Countries. Manchester: Chronic Poverty Research Center, 2001. ———. Thinking About Youth Poverty through the Lenses of Chronic Poverty, Life-Course Poverty and Intergenerational Poverty. Work document 57. Manchester: Chronic Poverty Research Center, 2005. Mora Salas, Minor, and Oliveira, Orlandina. "Las Vicisitudes de la Inclusión Laboral en los Albores del Siglo XXI: Trayectorias Ocupacionales y Desigualdades Sociales entre Jóvenes Profesionistas Mexicanos." Estudios Sociológicos 30, no. 88 (2012): 3-43. ———. "Los Jóvenes en el Inicio de la Vida Adulta: Trayectorias, Transiciones y Subjetividades." Estudios sociológicos 27, no. 79 (2009): 267-89. Moreno-Brid, Juan Carlos. "Crecimiento Económico y Transformación Productiva: Tres Décadas de Éxito Externo y Fracaso Interno ". In Informe del Desarrollo En México 2015, R. Cordera, and E. Provencio, México: PUED-UNAM, 2016. Moreno-Brid, Juan Carlos and Ros, Jaime. Desarrollo y Crecimiento en la Economía Mexicana. Una Perspectiva Histórica. Mexico: Fondo de Cultura Económica, 2010. Muthén, Bengt & Muthén, Linda. Mplus: Statistical Analysis with Latent Variables. Version 7.11 User's Guide. 2nd ed. Los Angeles, CA: Muthen & Muthen., 1998-2013. Navarrete, Emma Liliana. "Jóvenes y Trabajo: Un Reto por Resolver." In Trabajo y Derechos En México. Nuevas Afectaciones a la Ciudadanía Laboral, N. D´Angelo, M. Padrón Innamorato, F. Mancini, and L. Gandini, 147-66. Mexico: Universidad Nacional Autónoma de México, Instituto de Investigaciones Jurídicas, 2018. ———. "Problemáticas Laborales de México y Brasil en el Actual Contexto Económico." Estudios Demográficos 27, no. 1 (2012): 189-99. 60 Newton, Julie. Structures, Regimes, and Wellbeing. Work document of WeD 30, Bath: University of Bath, 2007. Nunnally, Jum C., & Bernstein, Ira H. Psychometric Theory. New York: McGraw Hill, 1994. Ochoa, Sara. "¿Es el Trabajo la Salida de La Pobreza? Los Trabajadores Pobres en México." In Cambio de Rumbo: Desafíos Económicos y Sociales de México Hoy, E. Vega López. México: PUED-UNAM, 2018. Ortega, María Elena and Ana Lucía Cárdenas. Mercado Laboral en México: Reporte para Informar el Diseño del Proyecto de Inclusión Laboral para Jóvenes. Manuscript, 2017. Ostrosky-Solís, Feggy & Lozano, Asucena. "Digit Span: Effect of Education and Culture." International Journal of Psychology 41, no. 5 (2006): 333-41. Parker, Susan. "Evaluación de Impacto de Oportunidades sobre Inscripción, Reprobación y Abandono Escolar." In Evaluación Externa de Impacto del Programa Oportunidades 2003, M. Hernández and B. Hernández, 13-95. Mexico, D.F: INSP, 2005. Parker, Susan and Behrman, Jere. "Seguimiento de Jóvenes Incorporados desde 1998 a Oportunidades: Impactos en Educación y Desempeño." In A Diez Años de Intervención. Evaluación Externa del Programa Oportunidades 2008 en Zonas Rurales (1997-2007), 199-238. Mexico, D.F.: INSP, 2008. Parker, Susan and Todd, Petra. "Conditional Cash Transfers: The Case of Progresa/Oportunidades." Journal of Economic Literature 55, no. 3 (2017): 866–915. Parker, Susan and Vogl, Tom. "Do Conditional Cash Transfers Improve Economic Outcomes in the Next Generation? Evidence from Mexico." NBER Working Paper, no. 24303 (2018). Parker, Susan; Behrman, Jere and Todd, Petra. "Impacto de Mediano Plazo del Programa Oportunidades sobre la Educación y el Trabajo de Jóvenes del Medio Rural que Tenían de 9 a 15 Años de Edad en 1997." In Evaluación Externa de Impacto del Programa Oportunidades 2004, M. Hernández and B. Hernández, 15-102. México D.F: INSP, 2005. Powers, Daniel A., and Xie, Yu. Statistical Methods of Categorical Data Analysis. United Kingdom: Emerald, 2008. Presidencia de la República. "“Decreto por el que se Crea la Coordinación Nacional del Programa de Educación, Salud y Alimentación como Órgano Desconcentrado de la Secretaría de Desarrollo Social”." Diario Oficial de la Federación (Federal Official Gazette), August 8, 1997. ———. "Sexto Informe de Gobierno. 1999-2000." Government of the Mexican United States. Mexico, 2018. Progresa. "Progresa: Programa De Educación, Salud y Alimentación." Poder Ejecutivo Federal. Mexico, 1997. Puga, Ismael and Solís, Patricio. "Estratificación y Transmisión de la Desigualdad en Chile y México. Un Estudio Empírico en Perspectiva Comparada." In Movilidad Social en México. Población, Desarrollo y Crecimiento, J. Serrano, and F. Torche, 189-228. Mexico: Centro de Estudios Espinosa Yglesias, 2010. Raferty, Adrian E. "Choosing Models for Cross-Classifications (Comment on Grusky and Hauser)." American Sociological Review 51 (1986): 145-46. Raudenbush, Stepehn W. "Comparing Personal Trajectories and Drawing Causal Inferences from Longitudinal Data." Annual Review of Psychology 52 (2001): 501−25. Raudenbush, Stepehn W., & Anthony S. Bryk. Hierarchical Linear Models: Applications and Data Analysis Methods. Advanced Quantitative Techniques in the Social Sciences Series. Second ed. Thousand Oaks, CA: Sage Publications, Ltd., 2002. Ridgeway, Greg; Kovalchik, Stephanie Ann; Griffin, Beth Ann, & Mohammed U Kabeto. "Propensity Score Analysis with Survey Weighted Data." Journal of causal inference 3, no. 2 (2015): 237-49. Ridgon, Edward E. "Cfi Versus Rmsea: A Comparison of Two Fit Indexes for Structural Equation Modeling." Structural Equation Modeling 3, no. 4 (1996): 369-79. 61 Robins, James M.; Hernan, Miguel Angel & Brumback, Babette. "Marginal Structural Models and Causal Inference in Epidemiology." LWW, 2000. Robles Berlanga, Héctor. "La Pequeña Agricultura Campesina y Familiar: Construyendo una Propuesta desde la Sociedad." EntreDiversidades 7 (2016): 46-83. Rodríguez-Oreggia, Eduardo and Freije, Samuel. "Movilidad Social Intergeneracional de los Jóvenes Beneficiarios de Oportunidades Provenientes de Hogares en Zonas Rurales." Manuscrito (2010). ———. "Una Evaluación de Impacto sobre el Empleo, los Salarios y la Movilidad Ocupacional Intergeneracional del Programa Oportunidades." In A Diez Años de Intervención. Evaluación Externa del Programa Oportunidades 2008 en Zonas Rurales (1997-2007), 61-124. México, D.F.: INSP, 2008. Roemer, John. "Defending Equality of Opportunity." The Monist 86, no. 2 (2003): 261–82. Rosenbaum, Paul R. Design of Observational Studies. 2nd ed. New York, Inc: Springer Verlag, 2009. Rubin, Donald B. "The Use of Propensity Scores in Applied Bayesian Inference." Bayesian statistics 2 (1985): 463-72. Samaniego, Norma. "Apertura Comercial y Trabajo en México." In Cambio de Rumbo: Desafíos Económicos y Sociales de México Hoy, E. Vega López. México: PUED-UNAM, 2018. Saraví, Gonzalo. Transiciones Vulnerables. Juventud, Desigualdad y Exclusión en México. México: Centro de Investigaciones y Estudios Superiores en Antropología Social, 2009. Sedesol y Oportunidades. "Rediseño del Programa de Desarrollo Humano Oportunidades: Consolidación de la Inversión en Capital Humano y su Impulso Hacia un Sendero de Mayor Prosperidad." Manuscript, 2014. Sen, Amartya. Poverty and Famines. An Essay on Entitlement and Deprivation. Oxford: Clarendon Press, 1981. Sewell, William; Haller, Archibald and Portes, Alejandro. "The Educational and Early Occupational Attainment Process." In Social Stratification in Sociological Perspective. Class, Race and Gender, D. Grusky, 410-21. Boulder: Westview Press, 2001. Sewell, William H.; Haller, Archibald O. and Ohlendorf, George W. "The Educational and Early Occupational Status Attainment Process: Replication and Revision." American Sociological Review 35, no. 6 (1970): 1014-27 Shavit, Yossi and Müller, Walter. From School to Work. Acomparative Study of Educational Qualifications and Occupational Destinations. Oxford: Clarendon Press, 1998. Solís, Patricio. "Cambio Estructural y Movilidad Ocupacional en Monterrey, México." Estudios Sociológicos 23 (2005): 43-74. ———. Inequidad y Movilidad Social en Monterrey. Mexico: El Colegio de México, 2007. ———. "Movilidad Intergeneracional de Clase en América Latina: Una Perspectiva Comparativa." In Y Sin Embrago se Mueve... Estratificación Social y Movilidad Intergeneracional de Clase en América Latina, P. Solís and M. Boado, Mexico: El Colegio de Méxio, Centro de Estudios Espinosa Yglesias (CEEY), 2016. ———. "Ocupaciones y Clases Sociales en México." In Movilidad Social en México: Población, Desarrollo y Crecimiento, J. Serrano and F. Torche. México: Centro de Estudios Espinosa Yglesias (CEEY), 2010. Solís, Patricio and Cortés, Fernando. "La Movilidad Ocupacional en México: Rasgos Generales, Matices Regionales y Diferencias por Sexo." In Tramas Familiares en el México Contemporáneo. Una Perspectiva Sociodemográfica., C. Rabell, 395-433. Mexico: UNAM-El Colegio de México, 2009. Solís, Patricio; Cortés, Fernando and Escobar, Agustín. "Discusión General." In Cambio Estructural y Movilidad Social en México, F. Cortés, A. Escobar, and P. Solís, 371-88. Mexico: El Colegio de México, 2007. 62 Sorensen, Aage. "The Basic Concepts of Stratification Research: Class, Status and Power." In Social Stratification in Sociological Perspective. Class, Race and Gender, D. Grusky, 287-300. Boulder: Westview Press, 2001. Sorensen, Annemette. "Welfare States, Family Inequality, and Equality of Opportunity." Research in social stratification and mobility 24 (2006): 367-75. Steiger, James H., & Lind, J. C. "Statistically Based Test for the Number of Common Factors " Paper presented at the Annual Meeting of the Psychometric Society, Iowa City, IA., 1980. Todd, Petra; Gallardo, Jorge; Behrman, Jere; and Parker, Susan. "Impacto de Oportunidades sobre la Educación de Niños y Jóvenes de Áreas Urbanas Después de Un Año de Participación en el Programa." In Evaluación Externa de Impacto del Programa Oportunidades 2004, M. Hernández, and B. Hernández, 165-227. Mexico, D.F: INSP, 2005. Torche, Florencia. "Cambio y Persistencia en la Movilidad Intergeneracional en México." In Movilidad Social en México. Población, Desarrollo y Crecimiento, J. Serrano, and F. Torche, 71-134. México: Centro de Estudios Espinosa Yglesias, 2010. Treiman, Donald, & Ganzeboom, Harry. "The Fourth Generation of Comparative Stratification Research." Research Council of the International Sociological Association. Montreal, 1998. Tucker, Ledyard R., & Lewis, Charles. "A Reliability Coefficient for Maximum Likelihood Factor Analysis." Psychometrika 38 (1973): 1-10. Turner, Bryan. Equality. Chichester: Ellis Horwood, 1986. Valdés , Servando and Vargas, Delfino. "Una Aplicación del Análisis de Clases Latentes para Formación de Tipologías de Rezago Social en México 2015." In Memorias del XXVII Simposio Internacional de Estadística 5th International Workshop on Applied Statistics. Medellín, Antioquia, Colombia, August 8 to 12, 2017. Valencia, Enrique; Foust, David; and Tetreault, Darcy. Sistema de Protección Social en México a Inicios del Siglo XXI. Santiago de Chile: CEPAL, 2012. Vargas, Delfino, & Lorenz, F. O. "Inference with Missing Data Using Latent Growth Curves." Revista del Instituto Interamericano de Estadística 67, no. 188 and 189 (2015): 9-19. Wong, Raymod Sin-Kwok. "Understanding Cross-National Variation in Occupational Mobility." American Sociological Review 55 (1990): 560-173. World Bank. "Descripción de la Base de Datos de la Encuesta de Evaluación Encel 2017-2018." 1-67, 2018a. ———. "Descripción de la Base de Datos del Panel de Evaluación Encel 1997-2017." 1-13, 2018b. ———. "The State of Social Safety Nets." World Bank. Washington, DC, 2018c. ———. Nota Metodológica de los Estudios de Resultados de Largo Plazo de PROSPERA Programa de Inclusión Social. Zonas Rurales. Manuscrito, 2018d. ———. "Crecimiento del PIB Per Cápita (% Anual)." https://datos.bancomundial.org/indicador/NY.GDP.PCAP.KD.ZG?locations=MX&view=chart. N/D. Consulted on Novemebr 20, 2018. Xie, Yu. "The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables." American Sociological Review 57, no. 3 (1992): 380-95. Yamaguchi, Kazuo. "Models for Comparing Mobility Tables: Toward Parsimony and Substance." American Sociological Review 52 (1987): 482-94. Yaschine, Iliana. ¿Oportunidades? Política Social y Movilidad Intergeneracional en México. Mexico: El Colegio de México and Universidad Nacional Autónoma de México, 2015. Yaschine, Iliana. Progresa-Oportunidades-PROSPERA, Veinte Años de Historia. Manuscript, forthcoming. 63 Yaschine, Iliana and Ochoa, Sara. "La Política de Superación de la Pobreza en México 1997-2014." In Informe del Desarrollo en México 2015, R. Cordera and E. Provencio, 121-48. Mexico City: PUED- UNAM, 2016. Yaschine, Iliana; Hernández, Bernardo; and Urquieta, José. "The Evolving Anti-Poverty Agenda in Mexico: The Political Economy of Progresa and Oportunidades." In Conditional Cash Transfers in Latin America, M. Adato, and J. Hoddinott, 55-77. Baltimore: John Hopkins University Press, 2008. Yúnez, Antonio. "Introducción General." In Los Grandes Problemas de México. Economía Rural A. Yúnez, 11-22. Mexico: El Colegio de México, 2010. Zenteno, Rene and Solís, Patricio. "Continuidades y Discontinuidades de la Movilidad Ocupacional en México." In Cambio Estructural y Movilidad Social en México, F. Cortés; A. Escobar, and P. Solís, 123- 61. México: El Colegio de México, 2007. 64 Appendix A. Conformation of working samples Figure A.1 shows the process for defining the study’s working samples. ENCEL 2017 selected a sample of 15,457 youths to be interviewed; only 12,470 households of them could be interviewed. Subsequently, the 1997-2017 panel database, integrated by the World Bank, included 12,150 of these observations. This was the database that we took as an input to build our working samples for the study. Figure A.1. Conformation of working samples Source: Own elaboration based on the World Bank (2018a, 2018b) and ENCEL 1997-2017. At that moment, the following cuts were made: 1. Individuals whose age was not reported or who had inconsistencies with age were excluded. The sample was reduced to 9,752 observations. 2. The sample was cut to only include the age range of interest (18-35 years) and those whose household was eligible for the program in 1997 and/or whose household had been a beneficiary of the program between 1997 and 2017 were included. 9,659 observations remained. 3. We only retained youths who had information filled out on the youth questionnaire, given that the analysis required basic occupational information of youths present in this questionnaire. 5,101 observations remained. 4. Finally, the last two cuts to the database were made according to the needs of the two analyses. Since the analysis of occupational mobility requires information on 65 the occupational characteristics of the youth and the provider, the sample was reduced to 3,084 observations. For the analysis of the determinants of occupational attainment, only the occupational information of the youths is necessary, so the sample was composed of 3,310 observations. The following table shows the comparison of some socioeconomic characteristics among 1) the 9,659 youths for whom there is information available about their household and members, 2) the 5,101 who have information available from the youth questionnaire, and 3) the 3,310 from the working sample for the analysis of the determinants of occupational attainment (which includes those of the occupational mobility study). The data presented herein come from the database of members. We can see that the groups are similar in terms of their: average age, ethnic background, literacy and years of schooling. The main differences are that the working sample is composed of youths who are employed in the labor market, while in the others there is a small percentage that are currently only studying and a higher percentage that do not study or work. Since females have a lower employment rate, our study sample contains a lower percentage of females. Table A.1. Comparison of different samples of ENCEL 2017 ENCEL 2017 ENCEL 2017 Household Youth Work sample Indicator Questionnaire Questionnaire (n=3,310) (n=9,659) (n=5,101) Sociodemographic characteristics Average age 26.25 25.99 25.98 Females (%) 49.85 56.34 38.55 Indigenous (%) 29.41 26.84 26.13 Single (%) 32.93 33.54 38.88 Condition of activity and educational characteristics Only studies (%) 2.89 1.6 NA Studies and works (%) 2.91 2.32 3.66 Only works (%) 65.36 65.01 96.34 Does not study nor work (%) 28.84 31.08 NA Literacy (%) 96.74 96.37 96.92 Average years of schooling 8.98 9.24 9.27 Source: Own elaboration from ENCEL 2017.             66 Appendix B. Conformation of comparison groups according to intensity of PROSPERA’s treatment Table B.1 shows the results of the latent classes analysis to determine the number of latent classes of intensity of PROSPERA’s treatment. The results of the mutual independence models are presented with a single latent class, followed by the model with two latent classes, and then by the model with three latent classes. The criteria used for the selection of the model of two classes were a combination of the lowest value of BIC (3), with an entropy close to the unit resulting from the classification (4), the percentage of observations per class had to be greater than 5% (5), the probability of correct classification of observations in each class (6), and a significant Lo-Mendell-Rubin likelihood test (7). Taking into account that this last test rejects the hypothesis that fitting our model with 3 classes would have been better than that with 2, the slightly lower probabilities of correct classification, and the almost negligible improvement in entropy and the BIC of the model with 3 classes, we chose the model with 2 classes. Table B.1. Determinants of the number of latent classes of treatment intensity (1) (2) (3) (4) (5) (6) (7) Probability of LMR Class -2xlog-L BIC Entropy % in the class correct Test classification (%) 1 279,600 280,308                            2 255,827 257,229 0.945 49% - 51% 0.982 - 0.985 p< 0.001 3 245,174 247,270 0.952 30% - 37% 0.973 - 0.982 p< 0.759 Source: Own elaboration from ENCEL 1997-2017 and historical data of cash transfers from PROSPERA. Table B.2 presents the diagnostic analysis performed to evaluate the balance of the comparison groups after applying the propensity score method. We can see that with the original sample, before the use of the scores, the comparison groups registered differences in some of the pre- intervention variables used for the construction of the weights. With the pseudo sample generated by applying the weights, these differences are no longer observable, which allows us to assume that a good balance is achieved between both groups. 67 Table B.2. Difference of means of pre-intervention variables, according to intensity of PROSPERA’s treatment Difference of Mean means Higher Lower Bias Pre-intervention variables p>t intensity intensity (%) Panel A: Original Sample Proportion with indigenous head 0.34 0.31 6.5 0.071 Ratio with indigenous spouse of the head 0.33 0.3 7.4 0.041 Proportion with male head 0.95 0.92 11.4 0.002 Proportion spouse of the head male 0.04 0.06 -12 0.001 Age of head 42.86 39.08 30.4 0 Age of the head's spouse 37.43 32.72 41.2 0 Schooling of head 3.43 4.21 -33.1 0 Schooling of the head's spouse 3.38 4.43 -48.8 0 Eligibility score (2003 method) 3.56 2.63 89.5 0 Proportion with bad wall 0.13 0.14 -4.3 0.237 Proportion with ground floor 0.6 0.56 9.6 0.007 Proportion with bad roof 0.72 0.71 2.7 0.456 Proportion with electric light 0.74 0.74 -1.1 0.761 Proportion with piped water 0.41 0.39 3.9 0.279 Ln of monthly income per adult equivalent 5.93 5.97 -5.1 0.153 Panel B: Pseudo Sample Proportion with indigenous head 0.34 0.34 0 0.994 Ratio with indigenous spouse of the head 0.33 0.33 0 0.995 Proportion with male head 0.95 0.95 0 0.992 Proportion spouse of the head male 0.04 0.04 0 0.994 Age of head 42.86 42.86 0 0.995 Age of the head's spouse 37.43 37.43 0 0.998 Schooling of head 3.43 3.43 0 0.999 Schooling of the head's spouse 3.38 3.38 0 0.995 Eligibility score (2003 method) 3.56 3.56 0.1 0.978 Proportion with bad wall 0.13 0.13 0 0.999 Proportion with ground floor 0.6 0.6 0 0.991 Proportion with bad roof 0.72 0.72 0 0.998 Proportion with electric light 0.74 0.74 0 0.997 Proportion with piped water 0.41 0.41 0 0.999 Ln of monthly income per adult equivalent 5.93 5.93 -0.1 0.989 Panel C: Balance sheet quality indicators Bias Pseudo Mean Median p>chi2 R2 Original Sample 20.5 9.6 0.221 0.001 Pseudo Sample 0.01 0.01 0 0.999 Note: The multiple imputation method was used due to the missing data. Source: Own elaboration from ENCEL 1997-2017 and historical data of cash transfers from PROSPERA. Panel A and panel B of table B.3 show the average time and transfers, as well as the demographic characteristics of the groups of low and high intensity of PROSPERA’s treatment, respectively. 68 Table B.3. Characteristics of households in the low treatment intensity class Panel rounds Variables 97 98o 99m 99n 2000m 2000n 2003 2007 2017i Panel A: Low intensity of treatment Average bi-monthly periods 95 Average transfers (2010 pesos) 31,414 74,172 212,477 Number of households with presence of youths in selected 1,380 1,358 1,293 1,321 1,290 1,295 1,283 1,089 1,462 sample Average household size 5.03 5.25 5.08 5.24 5.1 5 4.95 4.52 4.06 Average youths in the household of primary school 0.88 1.01 1.07 1.11 1.17 1.18 1.17 0.71 0.28 age Average youths in the household of secondary school 0.33 0.36 0.36 0.4 0.4 0.42 0.59 0.75 0.25 age Average youths in the 0.37 0.38 0.33 0.35 0.29 0.29 0.32 0.55 0.56 household of high school age Average elderly at household 0.06 0.06 0.07 0.07 0.08 0.06 0.09 0.11 0.19 Panel B: High treatment intensity Average bi-monthly periods 101 Average transfers (2010 pesos) 62,078 121,386 285,552 Number of households with presence of youths in selected 1,334 1,342 1,269 1,295 1,283 1,286 1,287 1,146 1,393 sample Average household size 8.07 8.32 8.1 8.22 8.05 7.61 7.11 5.84 4.21 Average youth in the household of primary school 2.1 2.08 2.02 2.02 1.99 1.92 1.5 0.74 0.25 age Average youths in the household of secondary school 1.16 1.3 1.34 1.35 1.33 1.35 1.32 1.02 0.2 age Average youths in the household of high school age 0.7 0.96 0.93 1.05 0.91 0.85 1.22 1.23 0.49 Average elderly at household 0.08 0.09 0.09 0.09 0.1 0.1 0.1 0.1 0.18 Source: Own elaboration from ENCEL 1997-2017 and historical data of cash transfers of PROSPERA.   69 Appendix C. Theoretical models of intergenerational occupational mobility The theoretical models of association between origin and occupational destination are described here. Figure C.1 presents a graphic version of the specification of each one. Model 1. Complete independence. There is no association between origin and destination, the observations in this table occur randomly. This model is used as a reference. Model 2. Homogeneous diagonal. Occupational inheritance is the determining factor of the association and is located on the diagonal of the R × C table, the terms of the diagonal are denoted by dp, which correspond to the cells (i, i) for i = 1,2 ..., 6. This model assumes that intergenerational inheritance has the same strength in all strata of origin. Model 3. Diverse diagonal. It is very similar to the homogeneous diagonal, but specifies more than one parameter in the main diagonal and assumes that the strength of the association is specific for each parameter. In this case three parameters are defined in the diagonal: dp1 (for the stratum NM, cell (1,1)), dp23 (for the strata Commerce, HM, LMM and LMS, cells (i, i) for i = 2 ...,5) and dp6 (for the agricultural, cell (6,6)). This model assumes that intergenerational inheritance displays differentiated strength among the different strata of origin, according to what was just specified. Model 4. Symmetry (2 parameters) with homogeneous diagonal. Homogeneous occupational inheritance between all the strata is considered and a symmetry parameter that represents short- distance mobility (or barriers to long-distance mobility) of the same strength for all strata, either ascending or descending. The same restriction as model 2 is used and sim parameters are added that are equal for cells (i, i + 1) and (i + 1, i) for i = 1,2 ..., 5. Model 5. Symmetry (3 parameters) with homogeneous diagonal. It assumes occupational inheritance with the same strength for the different strata of origin (dp, corresponding to the cells (i, i) for i = 1,2 ..., 6), and short-distance mobility (or barriers to long-distance mobility) with differentiated strength between downward (sim1 for cells (i, i + 1)) and upward mobility (sim2 for cells (i + 1, i)) for i = 1,2 ..., 5. Model 6. Symmetry (4 parameters) with different diagonal. It combines the dp parameters from model 3 and the sim parameters from model 4. Therefore, this model assumes that there is intergenerational inheritance of differentiated strength according to strata of origin and short- distance mobility (or barriers to long-distance mobility) that has the same strength for all strata, either ascending or descending. Model 7. Symmetry (5 parameters) with diverse diagonal. It combines the parameters dp of model 3 and the parameters sim1 and sim2 from model 5. Therefore, this model assumes that there is intergenerational inheritance of strength that is differentiated according to the strata of origin and short-distance mobility (or barriers in long-distance mobility) with differentiated strength between upward and downward movements. Model 8. Corners (2 parameters) with homogeneous diagonal. It contains the dp parameters as defined in model 2. In addition, it has a parameter that models barriers with the same strength for long-distance downward mobility from the NM stratum (esq for the cells (i, i + 1), with i = 70 1,2) and for long-distance upward mobility from the agricultural stratum (esq for the cells (i + 1, i)) for i = 4,5). This model assumes that intergenerational inheritance is the same for all strata of origin and that barriers in long-distance mobility have the same strength at the extremes of stratification. Model 9. Corners (3 parameters) with homogeneous diagonal. This is a model similar to number 8, where the diagonal parameter, dp, is defined as model 2. In addition, there are two parameters that model the barriers for long-distance downward mobility from the Non-Manual stratum (esq1 only for the cells (i, i + 1), with i = 1,2) and long-distance upward mobility from the agricultural stratum (esq2 for the cells (i + 1, i) for i = 4,5). This model assumes that intergenerational inheritance is the same for all strata of origin and that barriers to long-distance mobility have differentiated strength at the extremes of occupational stratification. Model 10. Corners (4 parameters) with diverse diagonal. It contains the diverse diagonal parameters (dp1, dp23 and dp6) such as model 3, and the corners parameter, such as model 8. This model assumes that intergenerational inheritance is different according to the stratum of origin and that the barriers in long-distance mobility have the same strength at the extremes of stratification. Model 11. Corners (5 parameters) with diverse diagonal. It is similar to model 10 but two parameters are added in the corners. The diverse diagonal parameters remain the same as in model 3 (dp1, dp23 and dp6) and the corner parameters (esq1 and esq2) as in model 9. This model assumes that intergenerational inheritance is differentiated according to the stratum of origin and that the barriers in long-distance mobility have different strengths at the extremes of occupational stratification.     71 Figure C.1 Graphical representation of the restrictions imposed on the models 1. Independence 2. Homogeneous diagonal NM Commerce HM LMM LMS Agriculture NM Commerce HM LMM LMS Agriculture NM * * * * * * NM dp * * * * * Commerce * * * * * * Commerce * dp * * * * HM * * * * * * HM * * dp * * * LMM * * * * * * LMM * * * dp * * LMS * * * * * * LMS * * * * dp * Agriculture * * * * * * Agriculture * * * * * dp 3. Diverse diagonal  4. Simmetry (1p) with homogeneous diagonal NM Commerce HM LMM LMS Agriculture NM Commerce HM LMM LMS Agriculture NM dp1 * * * * * NM dp sim * * * * Comerce * dp23 * * * * Commerce sim dp sim * * * HM * * dp23 * * * HM * sim dp sim * * LMM * * * dp23 * * LMM * * sim dp sim * LMS * * * * dp23 * LMS * * * sim dp sim Agriculture * * * * * dp6 Agriculture * * * * sim dp 5.  Simmetry (2p) with homogeneous diagonal 6.  Simmetry (1p) with diverse diagonal NM Commerce HM LMM LMS Agriculture NM Commerce HM LMM LMS Agriculture NM dp sim1 * * * * NM dp1 sim * * * * Commerce sim2 dp sim1 * * * Commerce sim dp23 sim * * * HM * sim2 dp sim1 * * HM * sim dp23 sim * * LMM * * sim2 dp sim1 * LMM * * sim dp23 sim * LMS * * * sim2 dp sim1 LMS * * * sim dp23 sim Agriculture * * * * sim2 dp Agriculture * * * * sim dp6 7.  Simmetry (2p) with diverse diagonal 8. Corners (1p) with homogeneous diagonal NM Commerce HM LMM LMS Agriculture NM Commerce HM LMM LMS Agriculture NM dp1 sim1 * * * * NM dp esq * * * * Commerce sim2 dp23 sim1 * * * Commerce esq dp * * * * HM * sim2 dp23 sim1 * * HM * * dp * * * LMM * * sim2 dp23 sim1 * LMM * * * dp * * LMS * * * sim2 dp23 sim1 LMS * * * * dp esq Agriculture * * * * sim2 dp6 Agriculture * * * * esq dp 9.  Corners (2p) with homogeneous diagonal 10. Corners (1p) with diverse diagonal NM Commerce HM LMM LMS Agriculture NM Commerce HM LMM LMS Agriculture NM dp esq1 * * * * NM dp1 esq * * * * Commerce esq1 dp * * * * Commerce esq dp23 * * * * HM * * dp * * * HM * * dp23 * * * LMM * * * dp * * LMM * * * dp23 * * LMS * * * * dp esq2 LMS * * * * dp23 esq Agriculture * * * * esq2 dp Agriculture * * * * esq dp6 11. Corners (2p) with diverse diagonal NM Commerce HM LMM LMS Agriculture NM dp1 esq1 * * * * Commerce esq1 dp23 * * * * HM * * dp23 * * * LMM * * * dp23 * * LMS * * * * dp23 esq2 Agriculture * * * * esq2 dp6 Source: Own elaboration.   72 Appendix D. Description of the structural model and its fit measurements The parameters of the structural model are estimated via maximum likelihood. However, the present model includes both categorical and continuous variables, so that the weighted least squares means and variance adjusted (WLSMV) estimation method, implemented in the MPLUS 7.11 program (Muthen and Muthen, 1998-2013), are used. To estimate the model that contains missing values, the maximum likelihood method with complete information was used (FIML, Arbuckle, 1996), so there was no need to impute values and this method turns out to be efficient (Vargas and Lorenz, 2017). Three indices are used to evaluate the fit of the model: RMSEA, CFI and TLI. The root mean square error approximation (RMSEA) was useful for calculating the degree to which the proposed model fits the population reasonably well (Steiger, 1990, Browne and Cudeck, 1993). Values lower than 0.05 are desirable, but values between 0.06 and 0.08 are also acceptable, while values close to 0.09 or greater are undesirable. The RMSEA is recommended for confirmatory models and large samples (Ridgon, 1996). The CFI belongs to a category of incremental fit measures that compare a proposed model against the null model, called the comparative fit index (Bentler, 1990). These indices should exceed the recommended level of 0.90 to provide evidence on whether to accept the proposed model. The TLI was proposed by Tucker and Lewis (1973) and its range of possible values is not regulated. A good fit is observed when obtaining values greater than or equal to 0.90. 73 Appendix E. Construction of latent variables for the occupational attainment model Here, we describe the construction of the latent variables of the occupational attainment model, the dependent variable of the model, as well as the asset index that is used as a manifest variable for the latent variable of social origin. The reliability of constructed latent variables is measured by the omega composite reliability  (McDonald, 1999). An omega value greater than 0.7 is considered acceptable (Nunnally and Bernstein, 1994). As can be seen, our latent variables had heterogeneous values (0.45 to 0.92). The omega coefficient is calculated from the following expression: ∑ ∑ ∑ where are the standardized factorial loads and the standardized residual variances 1 . E.1 Youths’ social origin The latent social origin variable is constructed from the ISEI of the occupation of the youth’s provider,56 the schooling of both the father and the mother, when the youth was 14 years old, as well as a previously calculated index of the assets and services present in the household (see below). The schooling of the father and the mother show the most relevant factorial loads that build the social origin. The composite reliability is  = 0.60. The factorial loads are shown below. Social Origin loads Provider’s ISEI 0.271 Father’s schooling 0.656 Mother’s schooling 0.696 Index of assets 0.438 E.2 Cognitive skills The latent variables related to both the cognitive abilities of youths and their providers are measured with manifest variables through two cognitive tests: Raven's progressive matrices (Alderton and Larson, 1990, Bilker, et al., 2012) that it is used to measure abstract reasoning (the test applied is a reduced version) and the digit span (Ostrosky-Solís and Lozano, 2006) that measures short-term memory capacity. The composite reliability is  = 0.46 and 0.52, respectively. The factorial loads are shown below.                                                              56 See below the description of the variable youth’s occupational attainment for more detail on the ISEI that also applies to the provider’s ISEI variable. 74 Youth’s Cognitive Abilities loads Provider’s Cognitive Abilities loads Youth Raven 0.645 Provider Raven 0.575 Youth Digit Span 0.545 Provider Digit Span 0.511 E.3 Youths’ education The latent variable education is measured based on the youth’s years of schooling and three variables related to having attended different educational levels (secondary, high school and superior) and the type of school that the youth attended for each of them. The variables for type of school are categorized on an ordinal scale 0, 1 and 2. Where the value 0 indicates that they did not attend that year, 1 indicates that they attended a low-quality school and 2 indicates that they attended a school with higher quality. In the case of secondary and high schools, the classification was based on the scores of the PLANEA test according to school types. General schools were classified as being of higher quality, while the other types of schools were considered as being of lower quality level; the information of the ENCEL 2017 does not distinguish between public and private. In the case for tertiary level, the ENCEL only distinguishes between public and private; the private tertiary level schools were taken as being of better quality as, considering that in the regional educational contexts, private institutions tend to offer better service. The years of schooling and the type of tertiary level education are the manifest variables that have the greatest weight in the construction of this latent variable. The composite reliability is = 0.92. The factorial loads are shown immediately. Education loads Youth’s years of schooling 0.989 Secondary type 0.678 High School type 0.859 Superior type 0.903 E.4 Youths’ first occupation The youth’s first occupation was measured with the age at the first occupation and the ISEI of the first occupation. The composite reliability is  = 0.45. The factorial loads are shown below. Youths’ First Occupation loads First Age 0.544 ISEI First 0.534 75 E.5 Youth’s occupational attainment Occupational attainment is measured through the ISEI of the youth's current occupation, which is a manifest variable. The version of the ISEI that corresponds to the International Standard Classification of Occupations (ISCO) 2008 was used, which is constructed based on variables such as education, occupation and personal income that were validated for 198,500 men and women of diverse countries, using data from 2002 to 2007 of the International Social Surveys Program. The objective of the index is to minimize the direct effect of education on earnings and maximize the indirect effect of education on them via occupation. For the construction of the variable, the routine provided in the file "isqoisei08" was used (Ganzeboom and Treiman, 2012). The descriptive statistics of the youth's ISEI variable are shown below. Variable n Mean Dev. Est. Asymmetry Curtosis min max ISEI 3310 23.70 14.36 1.75 6.25 11.01 88.70 E.6 Index of assets and services The asset index is made up by variables that account for the characteristics of the housing (piped water, number of rooms, gas stove, floor material) and its assets (blender, electric light, radio recorder, refrigerator, television) in 1997. These variables are binary, where 1 indicates the presence and 0 the absence. The scores are obtained from a confirmatory factor analysis (CFA) for binary variables, so that a high index accounts for a better condition of the household (based on the characteristics of the housing and assets). The largest factorial loads are observed for the assets and the smallest for the possession of agricultural land. The composite reliability is  = 0.91. The factorial loads are shown below. Asset Index loads Tubing water 0.400 Number of rooms 0.626 Gas stove 0.915 Blender 1.000 Electric light 0.914 Radio recorder 0.574 Fridge 0.898 TV 0.919 Has animals for work 0.149 Agricultural land 0.042 Material floor 0.709 76 Appendix F. Output percentage tables In table F.1, the tables of output percentages of the subgroups of the youth are presented by sex, ethnic background, migratory status and intensity of PROSPERA’s treatment. Table F.1. Output percentages. Youths ages 18-35 years old from ENCEL 2017. Total, by sex, ethnic background, migratory status and treatment intensity. Source: Own elaboration from ENCEL 2017.   77 Appendix G. Results of selected log-linear and log-multiplicative models This Appendix presents the values of the parameters of the three-way log-linear models, as well as the  coefficients of their respective log-multiplicative models. Tables G.1, G.2 and G.3 contain the results of the models that analyze the differences by sex, migratory status and intensity of PROSPERA’s treatment, respectively. Table G.1. Log-multiplicative models for constant and non-constant fluidity. Youths from 18-35 years of age of ENCEL 2017. Comparison by sex. Log- Constant Non constant multiplicative* Value of Value of  Parameter OR P Sex OR P Model 2. Homogeneous diagonal Male 2.29 <0.001 dp 2.16 <0.001 0.79 Female 1.88 <0.001 Model 4. Symmetry (2p) with homogeneous diagonal Male 2.39 <0.001 dp 2.24 <0.001 Female 1.88 <0.001 0.76 Hombre 1.23 <0.001 sim 1.14 0.052 Female 1.01 0.999 Model 8. Corners (2p) with homogeneous diagonal Male 2.40 <0.001 dp 2.23 <0.001 Female 1.88 <0.001 0.76 Male 1.35 0.015 esq 1.17 0.064 Female 0.99 0.961 Model 9. Corners (3p) with homogeneous diagonal Male 2.45 <0.001 dp 2.25 <0.001 Female 1.87 <0.001 Male 0.62 0.264 esq1 0.81 0.448 0.74 Female 1.03 0.943 Male 1.51 0.002 esq2 1.23 0.025 Female 0.99 0.939 (*) the values of  for the log-multiplicative models do not refer specifically to any of the parameters dp, sim or esq. Source: Own elaboration from ENCEL 2017 (n = 3,084). 78 Table G.2. Log-multiplicative models for constant and non-constant fluidity. Youths from 18-35 years of age of ENCEL 2017. Comparison by migratory status. Log- Constant Non-constant multiplicative* Parameter OR Value of P Status OR Value of P  Model 2. Homogeneous diagonal Non migrant 2.48 <0.001 dp 2.22 <0.001 0.26 Migrant 1.26 0.082 Model 4. Symmetry (2p) with homogenous diagonal Non migrant 2.63 <0.001 dp 2.31 <0.001 Migrant 1.24 0.133 0.23 Non migrant 1.22 <0.001 sim 1.15 0.038 Migrant 0.95 0.7 Model 8. Corners (2p) with homogenous diagonal Non migrant 2.63 <0.001 dp 2.30 <0.001 Migrant 1.25 0.115 0.23 Non migrant 1.29 0.011 esq 1.16 0.073 Migrant 0.94 0.696 Model 9. Corners (3p) with homogenous diagonal Non migrant 2.66 <0.001 dp 2.32 <0.001 Migrant 1.26 0.092 Non migrant 0.91 0.781 esq1 0.79 0.398 0.25 Migrant 0.53 0.225 Non migrant 1.35 0.005 esq2 1.22 0.026 Migrant 1.03 0.874 (*) the values of  for the log-multiplicative models do not refer specifically to any of the parameters dp, sim or esq. Source: Own elaboration from ENCEL 2017 (n = 3,084). 79 Table G.3. Log-multiplicative models for constant and non-constant fluidity. Youths aged 18-35 years of the ENCEL 2017. Comparison by treatment intensity. Log- Constant Non constant multiplicative* Value Intensity Value Parameter OR of P of Tx OR of P  Model 2. Homogeneous diagonal Lower 2.09 <0.001 dp 2.04 <0.001 0.94 Higher 2.00 <0.001 Model 3. Diverse diagonal Lower 1.91 0.245 dp1 3.03 <0.001 Higher 4.07 <0.001 Lower 1.59 <0.001 dp23 1.61 <0.001 0.91 Higher 1.63 <0.001 Lower 2.95 <0.001 dp6 2.65 <0.001 Higher 2.37 <0.001 (*) the values of  for the log-multiplicative models do not refer specifically to any of the parameters dp, sim or esq. Source: Own elaboration from ENCEL 2017 (n = 3,084). 80 Appendix H. Results of fitted structural models Here we present the process of selecting the general model for the analysis of the determinants of occupational attainment. Four models were estimated. Model 1 fits all the paths in Figure H.1 and the coefficients β11, β10, β4 and β9 are not significant. Models 2, 3 and 4 gradually discard non-significant paths, with which model 4 fits a parsimonious model, without substantively changing the coefficients adjusted by previous models (see Table H.1). On the other hand, fit indices remain almost intact when non-significant paths are discarded. For example, the RMSEA is 0.37 and 0.35, the CFI remains almost the same 0.978 and 0.974, in models 1 and 4, respectively. Model 4, which is the selected one, indicates that social origin does not directly affect the youth’s first occupation or current occupational  attainment. Likewise, the provider's cognitive skills do not have a direct effect on the youth’s education or occupational attainment. Figure H.1 Proposed structural model Source: Own elaboration. 81 Table H.1 Results of fitted structural models Model 1 Model 2 Model 3 Model 4 path coeff. coeff. coeff. coeff.  0.4 *** 0.4 *** 0.4 *** 0.4 ***  0.5 *** 0.5 *** 0.5 *** 0.5 ***  0.56 *** 0.56 *** 0.55 *** 0.55 ***  0.002 ns 0.002 ns 0.001 ns  0.69 *** 0.69 *** 0.7 *** 0.7 ***  0.09 ** 0.07 ** 0.07 ** 0.08 ***  0.14 ** 0.16 *** 0.16 *** 0.16 ***  0.32 *** 0.32 *** 0.32 *** 0.32 ***  0.04 ns 0.01 ns 0.01 ns  -0.01 ns -0.02 ns  -0.04 ns phi 0.66 *** 0.66 *** 0.66 *** 0.66 *** RMSEA 0.037 0.037 0.036 0.035 CFI 0.978 0.978 0.979 0.979 TLI 0.971 0.971 0.973 0.974 Levels of significance *** p <0.001, ** p <0.01, * p <0.05. Source: Own elaboration from ENCEL 2017 (n = 3,310).   82