The impacts of teacher working conditions and human capital on student achievement: evidence from brazilian longitudinal data

ABSTRACT

We use a new longitudinal matched student-teacher data from Brazil to examine the impacts of teacher human capital and working conditions on student learning controlling for student, teacher, and student-teacher match fixed effects. We find that teacher working conditions impact students outcomes. Teachers employed in more than one school negatively affect student test scores, while those working more hours in the same school positively impact students outcomes. Additionally, teacher education, experience, and family income affect student outcomes in Portuguese but not in Mathematics. Students who change from a low to a high value-added teacher experience a significant increase in proficiency.

KEYWORDS:

• Education
• working conditions
• student-teacher match
• fixed effects
• GERES

JEL CLASSIFICATION:

• I21
• I29

Acknowledgments

The authors acknowledge the financial support of FAPEMIG, CNPq, and CAPES-INEP (Observatório da Educação 2010). We are also grateful to CAED/UFJF for providing us with the GERES database, and data technical support of ECONS - Economics Research Laboratory from UFJF. Any errors are our own.

Data-availability

The data that support the findings of this study are available from the CAED/UFJF. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the members of the project at https://laedpucrio.wordpress.com/projetos/o-projeto-geres/ with the permission of them.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Several studies show that test scores are strongly related to success in the labor market (Leibowitz 1974; Murnane, Willett, and Levy 1995; Neal and Johnson 1996; Keane and Wolpin 1997; Cameron and Heckman 1998; Hanushek and Rivkin 2006a; Cunha and Heckman 2007, 2009).

² See Franco, Brooke, and Alves (2008) for a description of the data.

³ Rockoff (2004) uses a random-effects meta-analysis approach to measure the variance in teacher fixed effects accounting for measurement error and identifying significant effects of teachers and experience on vocabulary and reading.

⁴ The municipalities selected were those that presented the most adequate conditions for developing the analysis, including the proximity to the universities that took part in the study (Brooke and Bonamino 2011).

⁵ Schools that had only multi-grade classes for first grade were excluded from the sample. Private schools with fewer than 10 students enrolled in the second grade were also excluded, as were schools with fewer than 20 students. Furthermore, schools in rural zones and who had first grade students only at night classes were excluded. Thus, schools were selected based on their average socioeconomic level of the schools and their average size.

⁶ Tests were administered to students present in class. Therefore, differences in the samples for Portuguese language and Mathematics may arise. Some of the loss of students over time may be due to school transfers.

⁷ In Brazil, the K–12 system (basic education) is divided into two levels: primary education and high school. Primary education is compulsory and it is divided into Elementary Education I (1^st to 5^th grade) and Elementary Education II (6^th to 9^th grade). High school corresponds to grades 10–12. In this paper, our focus is on students from 1^st to 5^th grade, i.e. during the first cycle of primary education (Elementary Education I).

⁸ As the students from Salvador did not participate in all waves, they were excluded from our sample.

⁹ The regular age to attend Elementary Education I in Brazil is between 6 and 7 years old in the first year of elementary school and between 9 and 10 years old in the fourth year of elementary school. That means, the GERES project followed students that were 6–7 years old in 2005 (when the longitudinal panel started).

¹⁰ The majority of students are attending at a regular age and 13% are repeating students and thus, older than the others.

¹¹ To test whether schools are roughly the same in our sample, we decomposed the R² and we found that schools explain only 1.66% of the variance for Portuguese scores, and only 2.61% for Mathematics.

¹² We normalized the test scores across all students and waves. As previous explained, proficiencies are recalculated each wave, allowing us to compare different waves.

¹³ For more details, see Cameron and Trivedi (2010).

¹⁴ We estimated all models without grade dummies and obtained similar results.

¹⁵ The interpretation of the GERES proficiency scale is based on a range of skills development, from basic knowledge and abilities to more complex ones in Portuguese, and in Mathematics as well. See Brooke and Bonamino (2011) to a detailed explanation about the interpretation of the proficiency levels in the lower grades of elementary school.

Additional information

Funding

This work was supported by Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES-INEP Observatório da Educação, 2010).;;;