Bolsa Família Program in Brazil: assessing the impact on educational indicators of children and adolescents by regions1. Julio Racchumi2 Regiane Carvalho 3

Abstract The objective of this study is to evaluate the impact of receiving Bolsa Familia Program (BFP) on the school performance of Brazilian children and adolescents in different age groups and in different regions of the country. The BFP, created in 2003, benefits about 13 million families in poverty and extreme poverty through direct transfer of income. One of the conditions to receive the benefit is that children between 6 and 17 years old attend schools regularly. The data comes from two sources: an external evaluation of the Center for Public Policy and Education Evaluation (CAEd) applied to students in public education and the National Household Sample Survey (PNAD) conducted by the Brazilian Institute of Geography and Statistics (IBGE) both in 2011. After a simulation to estimate the beneficiary families in PNAD, the methodology employed is the Propensity Score Matching that allows comparison of beneficiary families with not beneficiaries ones whose observable characteristics are similar. The results illustrate, by some indicators, the impact of Bolsa Família on the school performance of children and adolescents after nine years of program deployment.

Key words: Education; Bolsa Familia Program; public policy; propensity matching.

1

XXVII IUSSP – International Population Conference. Busan, South Korea – August 2013. Centro de Políticas Publicas e Avaliação da Educação – CAEd/ UFJF/ Brazil – [email protected] 3 Centro de Desenvolvimento e Planejamento Regional – Cedeplar/ UFMG/ Brazil – [email protected] 2

1

Bolsa Familia Program and public policy in Brazil Brazil's main policy of conditional income transfer program is Bolsa Familia Program (BFP), created in 2003 from the unification of a number of other social benefits. In 2011, Bolsa Familia Program assisted more than 13 million families nationwide (BRAZIL, 2012), benefiting families in poverty and extreme poverty through direct income transfer in order to ensure the human right to adequate food, promoting nutrition security and contributing to the achievement of citizenship for the most vulnerable population. In the international context the programs of transfer income emerged in different ways, both in developed and in developing countries. In most developed region, these programs had goals as achieving and consolidating the Welfare State. The Nordic countries of Europe in over six decades were the pioneers in introducing income transfer programs, such as the United Kingdom in 1948, Finland in 1956, Sweden in 1957, Germany in 1961, Netherlands in 1960, Belgium in 1974 and in the case of France in 1988 (Rocha and Zimmermann, 2010). Latin America countries’, conditional income transfer programs are generally identified as a new generation of policies against poverty and misery (Zepeda, 2008). These programs, unlike the compensatory and assistentialist policies of earlier decades, aim the overcoming of social problems and poverty. Another feature of these programs is to focus benefits, which is justified by the need to increase the effectiveness of social spending, allocating scarce resources in more deprived groups. Mexico is one of the pioneers in adopting measures to combat poverty through the creation in 1997 of Progresa (currently OPORTUNIDAD), supporting families in rural and extremely poor areas aiming to improve the educational level, health status and nutrition of poor families specially children and mothers. In Colombia, Familias en Acción (FA) is a program that started in 2001 and aims to increase investment in human capital among poorest families, but also act as a social safety net. The Bolsa Familia Program operates from three main principles: direct income transfer, conditions and complementary programs. The direct income transfer promotes the immediate relief of poverty. The amounts of benefits paid by the BFP vary between R$32.00 to R$306.00 (US$14.00 to US$133.00) according to characteristics of each family - considering the family's monthly per capita income limited to R$140.00 (US$60.00), the number of children and adolescents up to 17 years, pregnant women and nursing mothers among family's components (BRAZIL, 2012). Conditions are the counterpart of the families receiving the program to reinforce their autonomy and access to basic social rights in the areas of education, health, employment and income generation. 2

One of the program’s conditions is that children and adolescents between 6 and 15 years of age are enrolled and attending school with a monthly minimum 85% of the workload. Students between 16 and 17 years must have frequency at least 75%. Thus, it is expected that beneficiaries have also gain in education indicators. The complementary programs aim the development of families so that beneficiaries can overcome the situation of vulnerability. These programs cover different areas such as education, work, culture, microfinance, training and improvement of housing conditions (BRAZIL, 2012a). Several studies have demonstrated the program’s irrefutable contribution regarding its short-term goal, namely, the reduction of social inequalities and poverty. The 4 th National Monitoring Report on the Millennium Development Goals, for example, showed a decrease in extreme poverty from 12% in 2003 to 4.8% in 2008 (IPEA, 2010). Therefore it isn’t the aim of this paper to question or emphasize the contribution of Bolsa Familia in short terms. However, it is expected that in medium and long terms it will be possible to observe other positive consequences of the program enabling beneficiary families to overcome the program dependence and to achieve autonomy on their own keeping themselves out of poverty. The aim of this study is to evaluate the impact on Brazilian students’ school performance whose families receive Bolsa Família Program. The specific aims are: compare school proficiency in Mathematics and Portuguese of children who receive and don’t receive Bolsa Família. Furthermore, controlling for familiars and household characteristics, test the impact of receiving BF on indicators of school performance: literacy, school attendance, and school delay. These exercises are made for age groups which 6 to 14 and 15 to 17 years old corresponding the cycles of basic education in Brazil. The results are evaluated for the whole country and different states and geographic regions. It is expected that, after nine uninterrupted performance years of the program, we can verify that the program provided the necessary support to achieve its secondary objective. We started from the hypothesis that receiving Bolsa Família exerts a differential impact by age group because the school attendance was practically universalized from 7 to 14 years of age, even before the program implementation. Therefore, it is expected that a greater impact is observed in the age group of 15-17 years old. It is also assumed that the program's impact is greatest in the poorest regions of the country, and it is expected that school performance is also affected by characteristics in family arrangements to which children and adolescents belong.

3

Bolsa Familia and Education Studies assessing the impact of a program, in general, assess the efficiency and effectiveness of programs and policies on the well-being of individuals (Patton, 2002; Cohen and Franco, 1988). This type of assessment measures the magnitude of the changes generated and the causality with the components and the benefits granted by the interventions. Policies and programs correspond to the causes, the effects are all changing conditions of the beneficiaries measured as changes attributable to the intervention (Heckam and Vytlacil, 2005). In Brazil, several studies have been published in the last decade regarding Bolsa Familia especially researches on the impacts of receiving the benefit in fulfilling their conditions and poverty reduction (Oliveira et al., 2007; Rocha, 2011). The conditionality of income transfers in exchange for school attendance is justified by the assumption that poor families would have high costs to keep their children in schools due to low and unstable family income, causing the need for early work to compete with continuous studies. In addition, low education of those responsible for poorer families would be a determining factor for the low pay. Thus, with no income transfers, the vicious circle of poverty would be repeated every generation, because hardly the current generation could break this circle and achieve autonomy by itself. So education would be a form of social inclusion and ensure the poorest families the opportunity to acquire social capital and thus subsidize their social mobility in the medium and long term. Assess the success extent due to the conditionality of receiving Bolsa Familia Program on education is of paramount importance because the returns of formal education are high, been one of the most efficient and sustainable ways to break the circle of poverty. The analysis importance of disaggregated coverage levels is justified by the need for better evaluation and planning of public policies given the deep regional inequalities existing in Brazil.

Data and methodology The first database used is an evaluation from Center for Public Policy and Education Evaluation (CAEd) held in 2011. This evaluation was carried in public schools in some Brazilian states and focuses on the teaching in the 5th and 9th level of elementary school and 3rd level of high school. With this basis was made a comparison of proficiency mean in two basic disciplines: Mathematics and Portuguese language considering students receiving or not BFP. 4

The second database used is PNAD (National Household Sample Survey) 2011, conducted by the National Institute of Geography and Statistics (IBGE). This is a nationwide annual survey based on a probability sample into three stages: municipalities, census tracts and households. In 2011, the survey investigated issues such as education, migration and fertility. However, there isn't an exclusive question in the National Household Survey about enrollment in Bolsa Família Program for any member of the household. There is only the question about "other income" (V1272), which may include, in addition to social benefits, financial investments. Thus, from a methodology developed by Souza et al. (2011), four procedures were adopted to make the PNAD 2011 adequate to the purposes of impact evaluation: 1. Data cleaning; 2. Disaggregation of income related to BFP; 3. Compatibility of the number of families served by this program in the National Household Survey and administrative records; 4. Reclassification and exclusion of families with zero income, but with no poor profile. The cleaning of the database consisted in exclude residents who did not belong to the family group (boarder or domestic servant) and also those who had any ignored income. This cleansing of the data resulted in excluding 2.5% of the people. Then the families who reported receiving income from social programs up to R$306.00 (maximum amount payable by the program) were identified as beneficiaries of BFP. With this estimate 10.4 million cases were obtained. However, according to administrative records, there were 13.3 million beneficiary households in 2011 (BRASIL, 2012b). Given this underestimation, it was necessary to estimate the probability of non-beneficiary families inclusion. Then, a matching was made to randomly select 2.9 million missing families. Finally, using a statistical analysis of conglomerate and based on a number of variables, a selection of the families that had zero income in the month of the survey but didn’t have the profile of the poor families was made. Then, these families were removed not to bias the sample. Following, to identify beneficiary families in PNAD, it was used the method of propensity score matching to test the impact of receiving Bolsa Familia on the five indicators of schooling: School attendance for children aged 6 to 14 years, school delay for children aged 6 to 14 years, school attendance for young aged 15 to 17 years, school delay for young aged 15 to 17 years and illiteracy of children and young aged 10 to 17 years. The results were evaluated for both sexes and for the five major geographical regions of Brazil, as well as for the whole country. The propensity score method takes the characteristics of the beneficiaries as a basis, in order to find, in the comparison group, non-beneficiaries who have these same characteristics (Rosenbaum Rubin 5

1983; Imbens 2000). The propensity score is the probability of a family or household receive the benefit of Bolsa Familia Program. The objective of matching is to find, among non-beneficiaries, an ideal comparison group to compare to the group of beneficiaries. Beneficiaries and non-beneficiaries families were matched considering a set of observable variables related to the household head, also in relation to their own household. In relation to the household head, characteristics such as sex, race / color (white or nonwhite), education (years of schooling) and migration (residence in the city less than 5 years) were considered. Regarding the household, were considered the location of the household (urban or rural), housing quality, considering the structure of the household (predominant material on the walls, floor and ceiling), access at home to basic services (sanitation, electricity, potable water and bathroom for exclusive use) and durable goods (television, computer, washing machine, etc.), proportion of children under 14 and elderly people over 60 in the household, type of family arrangement and total residents. Once calculated propensity scores, the next step is to use the matching to define what are the control elements for each beneficiary family, allowing the calculation of the average treatment effect (ATT). This effect is estimated by the difference between the results for the group of beneficiaries and for the matched comparison group. To confirm the "propensity score matching" results consistency, the technique of "nearest neighbor" without replacement in three different models was used.

Results The following are the results of comparison of proficiency among beneficiaries and nonbeneficiaries students and the results of impact evaluation of receiving Bolsa Família on the selected educational indicators. 

Descriptive results

Figures 1 and 2 show the difference in the average proficiency of students in Mathematics and Portuguese language for elementary and high school, respectively. For all states analyzed, it was observed that beneficiaries of BFP, of both sexes, reach lower average proficiency than non beneficiaries students. This is a result that indicates that only receiving the benefit is not enough for children who are in school to acquire the necessary proficiency as expected.

6

Figure 1. Average proficiency of elementary school students, according benefit of Bolsa Familia Program and discipline assessed. Brazil, 2011.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011.

Figure 2. Average proficiency of high school students, according benefit of Bolsa Familia Program and discipline assessed. Brazil, 2011.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011. 7

Controlling the average proficiency by mother's education (results in Appedix), we also observed that the beneficiaries continue presenting proficiency lower than non beneficiaries. The proficiency between beneficiaries and non-beneficiaries is more similar among children of mothers with low education, which could demonstrates that the program is focused on people with worse conditions making potential impacts more visible in these cases, equaling beneficiaries and non-beneficiaries. 

Impact analysis of receiving Bolsa Família

The impact analysis ensures that beneficiaries are being compared with non-beneficiaries who have the same or very similar family group characteristics observed. Table 1 shows only the statistically significant results in at least one of the tested models. The final Table with all the results is in the Appendix. ATT results are interpreted in terms of percentages and are presented by gender for the whole country and its the four major administrative regions (North, Northeast, Southeast and South), since the results of the Midwest were not significant. Brazil: For men, the results were significant indicating that between beneficiaries there is a lower percentage of children (6-14 years) delayed compared with non beneficiaries with similar family characteristics. For women the results were significant for children between 6-14 years, indicating that those who receive the assist of Bolsa Familia have higher school attendance and less school delay compared with children who don’t receive the benefit and have similar family characteristics. Among female young group, those who receive Bolsa Family have more school delay than those who don’t receive the program. There were also significant and positive results with regard to literacy and young children aged 10 to 17 years. Northern: For male children aged 6 to 14 years we find that beneficiaries of Bolsa Familia Program have higher school attendance and less school delay compared those who don’t receive the benefit. For young women the results were negative in relation to school delay meaning that, in this group, beneficiaries have more school delay than non beneficiaries. At same time, beneficiaries aged 10 to 17 years old are less illiterate than non-beneficiaries. Northeast: is the poorest region of the country and where we expected to find the more significant results, but this didn’t happened. For male students we find no significant results. For female students, we find that girls aged 6 to 14 years presented positive and significant indicators of school attendance and school delay. The school attendance for the beneficiaries was 2.1% higher and the school delay was 1.8% lower than non beneficiaries.

8

Table 1 - Average treatment effect on the treated (ATT) - Brazil and regions by sex, 2011. Nearest neighbor Dependent variables

nd=5 ATT

Brazil Male Children without school delay aged 6 to 14 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Northern Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years Female Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Northeast Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years Southeast Male Children without school delay aged 6 to 14 years Female Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years South Female School attendance of children aged 6 to 14 years Young without school delay aged 15 to 17 years

nd=8 T-ST

ATT

nd=10 T-ST

ATT

T-ST

0,0187

2,1605** 0,0174

2,2055** 0,0189

2,5049**

0,0158 0,0203 -0,0150 -0,0111

1,8212*** 2,6453*** -2,4546** -2,2382**

1,9949** 2,3902** -2,5825** -2,2679**

0,0160 0,0161 -0,0125 -0,0083

2,1561** 2,3888** -2,2403** -2,3482**

-0,0317 0,0285

-1,0642 0,0208 2,0089** 0,0263

1,3964 1,9552*

0,0280 0,0268

1,9270* 2,0353**

-0,0213 -0,0108

-1,3945 -1,6971*

-0,02442 -1,7107* -0,0230 -0,012166 -2,2091** -0,0103

-1,6570* -1,9990**

0,0156 0,0202

1,1347 1,6483

0,0205 0,0180

0,0366

0,0154 0,0167 -0,0147 -0,0090

1,6246 1,5874

0,0214 0,0185

1,7485* 1,6780*

2,2033** 0,0292

1,8416*

0,0285

1,8272*

-0,0302 -0,0061

-2,6665*** -0,0210 -1,2744 -0,0069

-1,9690* -1,6454

-0,0201 -0,0068

-1,8811* -1,7801*

0,0484 -0,0257

2,1797** 0,0518 -1,7932* -0,0187

2,4944** 0,0522 -1,3859 -0,0149

2,5732** -1,1294

Bic

- to

-2,601

-1,972

-1,653

Uni

- to

-2,345

-1,653

-1,2860

0,01***

0,05**

0,10*

Source: microdata from PNAD 2011.

Alfa =

9

Southeast: The school delay for male beneficiaries aged 6 to14 years was about 3% lower. For young female beneficiaries we found about 2% more school delay than non-beneficiaries and less illiterate beneficiaries than illiterate non beneficiaries. South: For male students we found no significant results. Beneficiaries girls aged 6 to 14 years presented positive and significant indicators of school attendance, about 5% more than non beneficiaries. In other hand, beneficiaries girls aged 15 to 17 years present more school delay than non-beneficiaries. In short, the results were expected to verify a positive impact of receiving Bolsa Família Program. It is important to note that positive differences, although not statistically significant, were observed in school attendance in most of cases (Tabel 4 – Appendix), indicating slightly greater school attendance for children of BF compared with children non beneficiaries. Negative results for school delay indicator for young are somehow expected, considering that these young people were included in the program from 2008 and the school delay may have been acquired prior entering the program. However, contrary to expectations, this impact wasn’t greater in the poorest regions or positive for school attendance of young people 15-17 years. Furthermore, it is hard to notice any significance pattern between regions, or between sexes.

Final remarks Some considerations should be made on the database and the methodology employed in this paper. The use of a database that wasn’t designed to the proposed in this work was a limitation and a challenge. The PNAD isn’t a focused study to evaluate results in social programs and doesn’t even have specific variable to receipt of Bolsa Família. However, the use of this database was justified because it is the latest available at the time of the study, and the other recent bases available, for example, the Demographic Census of 2010 also were not suitable because it doesn’t have an exclusive question about receiving Bolsa Família Program. Other considerations should be made on the methodology employed. First, the results originate from a cross-sectional survey, i.e., at a single point in time, therefore, differences between beneficiaries and non-beneficiaries more than the impact of the policy are observed. Second, we must bear in our minds that the methodology of the propensity score only limits the bias of nonrandomness, for even matching by observable characteristics, however there are a number of other features - observable or not - that certainly impact on the treatment and control groups and changing the observed result. Other important point is that there isn’t a variable available to control the time 10

of receipt of the benefit, or even receive the benefit indeed, it can also cause some bias in interpreting the results. This study’s results show that in similar socioeconomic conditions, children receiving Bolsa Família have had higher school attendance, lower school delay and lower illiteracy than those not receiving the program. But for young people, the results, when significant, indicates a worse situation for the young beneficiaries of the program in relation to school delay indicator and not significant in relation to school attendance. In other words, as indicated by other research, the program achieves the goal of keeping children in school and reduces school delay, but when we analyze the average proficiency, beneficiary children have lower grades than non-beneficiaries ones. This indicates that more than fulfill the condition of school attendance, public policy must ensure that these beneficiary children have the necessary conditions to learn the content taught in the classroom ensuring the poorest families the opportunity to acquire social capital that will provide, in the medium and long terms, the means to overcome poverty. Regarding the non positive results for young people, one of the reasons may be the late entry of these youth in the program when they acquired school delay or even encouragement monthly financial data for young (R$38.00 or US$16.00) may not be as attractive as it is for the children. Regarding gender comparison, it wasn’t possible to observe a pattern of impact. For regions and age groups the results were not as expected, because it wasn’t among the poorest or young people we found the greatest program impact. Finally, whereas since its implementation there were two specific evaluations on BF, the first round of the research AIBF (Impact Evaluation of Bolsa Família) held in 2005 and the second in 2009, one recommendation of this work is the inclusion of a specific question about receiving Bolsa Família in PNADs that are held annually. Although it isn’t the purpose of this research, with the inclusion of only one question, it will be possible to, more accurately, use this database to carry out a series of studies on the beneficiaries of BF. What is justified considering that BFP is a policy that spends almost 30 million a year and its results have to be monitored continuously. Other recommendation of this paper is to increase further studies only with beneficiaries in lower income group to verify the significance of differences in receiving Bolsa Família to the poorest families making more specific valuation.

11

Bibliography BRASIL. Ministério de Desenvolvimento Social e Combate à Fome. O Programa Bolsa Familia. Brasília, DF, 2012a. BRASIL. Ministério de Desenvolvimento Social e Combate à Fome. Relatório de Gestão do exercício de 2011. Brasília, 2012b. CAED. Centro de Políticas Públicas e Avaliação da Educação. Sistemas de Avaliação Educacional nos Estados. 2011. COHEN E, FRANCO R. Evaluación de proyectos sociales. Santiago de Chile: Instituto Latinoamericano y Del Caribe de Planificación Económica y Social, 1988. HECKAM J, VYTLACIL E. Structural equations, treatment effects and econometric policy evaluation. Cambridge: National Bureau of Economic Research, 2005 (NBER Technical Working Paper, 306). IBGE. Banco de Dados. Pesquisa Nacional por Amostra de Domicílios de Minas Gerais – PNADMG. 2011. IMBENS GW. The role of propensity score in estimating dose: response functions. Biometrika, 87(3): 706-710. 2000. IPEA. Instituto de Pesquisa Econômica e Aplicada. Relatório Nacional de Acompanhamento. Objetivos de Desenvolvimento do Milênio. Brasília: Ipea, 2010, 184p. OLIVEIRA A et al. Primeiros resultados da análise da linha de base da pesquisa de avaliação de impacto do programa bolsa familia. IN: VAITSMAN, J.; SOUSA, RP. Avaliação de políticas e programas do MDS – Resultados: Bolsa Familia e Assistência Social. Brasília, DF: Ministério de Desenvolvimento Social e Combate à Fome, Secretaria de Avaliação e Gestão da Informação, 2007, v.2. PATTON M. Qualitative research and evaluation methods. 3ed. Thousand Oaks: SAGE, 2002. ROCHA S. O programa Bolsa Familia: evolução e efeitos sobre a pobreza. Economia e Sociedade, Campinas, v20, n1 (41), p. 113-139, abr. 2011. ROCHA S, ZIMMERMANN C. O Brasil e as Experiências Internacionais de Programas de transferência de renda. In 13o Basic Income as an Intrument for Justice and Place - BIEN Congress, 2010, São Paulo, Brasil.

12

SOUZA P, OSORIO R, SOARES S. Uma metodologia para simular o Programa Bolsa Família. Working paper 1654. IPEA, Brasília, 2011. ROSENBAUM P, RUBIN D. The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrica, 70(1): 41-50. 1983. ZEPEDA E. Transferências condicionadas de renda (TCR) reduzem a pobreza? One Pager, Brasília, n. 21, abr. 2008.

13

Appendix

14

Figure 1 - Average proficiency in Mathematics for elementary school students according to receiving BF and mother's education, by sex.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011.

Figure 2 - Average proficiency in Portuguese language for elementary school students according to receiving BF and mother's education, by sex.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011. 15

Figure 3 - Average proficiency in Mathematics for high school students according to receiving BF and mother's education, by sex.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011.

Figure 4 - Average proficiency in Portuguese language for high school students according to receiving BF and mother's education, by sex.

Source: Sistema de Avaliação Educacional do CAEd/Brasil. 2011. 16

Table 4 - Average treatment effect on the treated (ATT) - Brazil and regions by sex, 2011. Nearest neighbor Dependent variables

nd=5 ATT

Brazil Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Northern Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Northeast Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years

nd=8 T-ST

ATT

nd=10 T-ST

ATT

T-ST

0,0059 0,0187 0,0058 -0,0057 -0,0049

0,5952 2,1605** 0,4094 -1,0708 -0,6040

0,0103 0,0174 0,0144 -0,0061 -0,0009

1,1807 2,2055** 1,0765 -1,1816 -0,1354

0,0082 0,0189 0,0145 -0,0061 -0,0004

0,9928 2,5049** 1,1022 -1,1934 -0,0689

0,0158 0,0203 0,0082 -0,0150 -0,0111

1,8212*** 2,6453*** 0,6372 -2,4546** -2,2382**

0,0154 0,0167 0,0074 -0,0147 -0,0090

1,9949** 2,3902** 0,6133 -2,5825** -2,2679**

0,0160 0,0161 0,0108 -0,0125 -0,0083

2,1561** 2,3888** 0,9141 -2,2403** -2,3482**

-0,0317 0,0285 -0,0030 -0,0116 -0,0099

-1,0642 2,0089** -0,0843 -0,7938 -1,0690

0,0208 0,0263 0,0021 -0,0121 -0,0054

1,3964 1,9552* 0,0619 -0,8836 -0,6904

0,0280 0,0268 0,0014 -0,0152 -0,0062

1,9270* 2,0353** 0,0401 -1,1406 -0,8644

0,0179 0,0209 0,0260 -0,0213 -0,0108

1,1816 1,5464 0,7618 -1,3945 -1,6971*

0,0119 0,0150 0,0438 -0,02442 -0,012166

0,8460 1,1581 1,3694 -1,7107* -2,2091**

0,0094 0,0181 0,0469 -0,0230 -0,0103

0,6808 1,4248 1,5044 -1,6570* -1,9990**

0,0024 0,0196 0,0244 -0,0119 -0,0063

0,1514 1,4340 1,0661 -1,3424 -0,5857

0,0074 0,0152 0,0273 -0,0093 -0,0094

0,5260 1,1856 1,2811 -1,0931 -1,0317

0,0073 0,0177 0,0256 -0,0050 -0,0056

0,5376 1,4255 1,2217 -0,6014 -0,6605

0,0156 0,0202 -0,0030 -0,0025 0,0055

1,1347 1,6483 -0,1444 -0,2501 1,0629

0,0205 0,0180 -0,0028 -0,0034 0,0029

1,6246 1,5874 -0,1457 -0,3529 0,6673

0,0214 0,0185 0,0075 -0,0028 0,0024 Continues

1,7485* 1,6780* 0,3927 -0,2927 0,5971

17

Midwest Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Southeast Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years South Male School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years Female School attendance of children aged 6 to 14 years Children without school delay aged 6 to 14 years School attendance of young aged 15 to 17 years Young without school delay aged 15 to 17 years Illiteracy of children and young aged 10 to 17 years

Continuation 0,0160 0,0305 0,0302 0,0026 0,0006

0,6042 1,1887 0,6593 0,1336 0,0627

0,0322 0,0226 0,0326 0,0083 -0,0012

1,3160 0,9179 0,7381 0,4303 -0,1357

0,0307 0,0248 0,0457 0,0119 -0,0003

1,2801 1,0257 1,0417 0,6246 -0,0376

0,0138 0,0237 0,0066 -0,0231 -0,0018

0,5883 1,0811 0,1513 -1,0769 -0,3060

0,0076 0,0140 0,0079 -0,0149 -0,0017

0,3426 0,6602 0,1872 -0,7365 -0,3546

0,0070 0,0093 0,0097 -0,0122 -0,0023

0,3218 0,4452 0,2324 -0,6164 -0,4903

0,0255 0,0366 0,0222 0,0099 -0,0063

1,4372 2,2033** 0,8057 0,9971 -0,9000

0,0219 0,0292 0,0252 0,0052 -0,0068

1,3352 1,8416* 0,9508 0,5420 -1,1217

0,0229 0,0285 0,0255 0,0038 -0,0060

1,4346 1,8272* 0,9820 0,3996 -1,0047

0,0098 0,0184 0,0017 -0,0302 -0,0061

0,6232 0,0114 1,2912 0,0128 0,0710 -0,0009 -2,6665*** -0,0210 -1,2744 -0,0069

0,7822 0,9539 -0,0405 -1,9690* -1,6454

0,0099 0,0116 0,0069 -0,0201 -0,0068

0,6969 0,8732 0,3072 -1,8811* -1,7801*

-0,0039 0,0036 -0,0235 -0,0152 -0,0010

-0,1676 0,1671 -0,5565 -0,9662 -0,1595

0,0107 0,0043 -0,0210 -0,0205 -0,0039

0,4927 0,2093 -0,5273 -1,3512 -0,6632

0,0114 0,0073 -0,0248 -0,0176 -0,0031

0,5345 0,3602 -0,6331 -1,1475 -0,5616

0,0484 -0,0056 0,0533 -0,0257 0,0000

2,1797** -0,2904 1,4753 -1,7932* 0,0000

0,0518 -0,0088 0,0380 -0,0187 0,0003

2,4944** -0,4817 1,0850 -1,3859 0,0675

0,0522 -0,0049 0,0394 -0,0149 -0,0008

2,5732** -0,2699 1,1408 -1,1294 -0,2040

Bic

- to

-2,601

-1,972

-1,653

Uni

- to

-2,345

-1,653

-1,2860

0,01***

0,05**

0,10*

Alfa =

Source: microdata from PNAD 2011.

18