The Impact of School Infrastructure and Teachers’ Human Capital on Academic Performance in Brazil

Abstract

To improve one of the lowest rates of literacy and numeracy in the world, the government of Brazil has targeted public education reform, given the strong link between an educated population and economic growth. This study examines the academic performance of the Brazilian public primary school system. It addresses the empirical shortcomings of prior research to examine the dynamics of the relationship between academic performance scores and several demographic and institutional variables, such as socioeconomic characteristics, variations in school infrastructure and school complexity, and teachers’ human capital. We employed quantile regression to explore the determinants of academic performance across 35,490 schools in rural and urban environments in Brazil. The dependent variable in our analysis captures the academic performance score, as measured by Brazil’s education authorities, of each school in our dataset. The model includes several education-related indices used in prior research and, as explanatory factors, measures of teachers’ human capital and the students’ socioeconomic level, which synthesizes information on parents’ education and household income. The results suggest that several institutional variables, including access to school libraries, computer facilities, projectors, and televisions, are positively and significantly related to the academic performance of primary students in Brazil’s system of public education. Furthermore, students’ socioeconomic level is positively associated with their academic performance.

Keywords

primary education school infrastructure education expenditures teachers’ human capital

JEL Codes: I21, I28, I29

Introduction

According to the Organization for Economic Cooperation and Development (OECD, 2017), Brazil has one of the lowest rates of literacy and numeracy, only 35%.¹ Countries at the top of this ranking maintain rates at or higher than 85% (OECD, 2017). To tackle the problem that a study by Mbiti (2016) indicates persists across the developing world, the Brazilian government adopted the National Plan of Education (PNE) 2014–2024. The PNE has 10 targets related to the Índice de Desenvolvimento da Educação Básica (IDEB) (Basic Education Development Index) that focus on placing Brazil on the same educational level as developed countries (Alves & Soares, 2013). These targets, as analyzed by Almeida et al. (2013), include student access to schools, student performance, school quality, and teacher supply and compensation.²

The government’s focus on literacy and numeracy rates is understandable. The importance of an educated populace in stimulating national economic growth is well established in the economics literature.³ The critical role played by teachers in educating the population has received increasing attention. For example, Hanushek (2011) found that a teacher who is a single standard deviation above average in terms of pedagogical effectiveness produces marginal gains of more than $400,000 per year when working with a class enrollment of 20. For the U.S., this finding suggests that if the least effective 5%–8% of teachers in the U.S. were replaced with teachers of average pedagogical skill, the education system would rank first in STEM fields, yielding an economic benefit (i.e., the present value of the relative increase in world ranking) of $100 trillion (Hanushek, 2011). The benefits to a less developed country such as Brazil from a similar realignment would also be quite remarkable.

Other covariates that explain academic performance within and across a country’s public school system include socioeconomic characteristics, school infrastructure, and complexity.⁴ Although many studies have analyzed the different impacts of environmental conditions on school outcomes, none has analyzed how their impacts vary across both high- and low-performing schools, as well as across both urban and rural settings. The identification of differences between these types of comparisons is central to the formulation of more efficient education policies. Some studies have identified the determinants of IDEB results using logistic regression, while others have identified the influence of internal and external school factors through hierarchical linear regressions.⁵ However, both methodologies fail to explain how these factors influence schools with different IDEB results. We aim to find the answer through the utilization of a quantile regression approach.

This study extends the literature by investigating the different impacts of various demographic and institutional variables (e.g., socioeconomic characteristics, school infrastructure, school complexity, and teachers’ human capital) on academic performance for high- and low-performing schools as well as urban and rural schools. In doing so, we address the empirical shortcomings of prior research by employing quantile regression to examine the relationships between academic performance scores, measured using the IDEB scores,⁶ and data from 35,522 primary schools from the 2015 Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira (INEP) census. The results find a set of education infrastructure variables related to information technology and multimedia, sanitation, personnel support, and sports to be positively and significantly related to the academic performance of primary students in Brazil’s system of public education. Additionally, both teachers’ human capital, measured by college degree completion, and the socioeconomic level of parents, captured by synthesizing parents’ education and household income information, are positively associated with academic performance. However, these variables differently impact urban and rural schools with high- and low-academic performance. For example, while socioeconomic status is relevant for low-performing schools, the professional qualifications of teachers have a greater impact in high-performing ones. This dynamic changes for rural schools, where socioeconomic conditions are almost identical between schools with lower and higher performance; however, teachers’ education is four times higher in high-performing schools.

In the next section, we provide a review of prior literature, with particular attention to published work on the relationship between education infrastructure, teachers’ human capital, and the academic performance of primary students. The Methodology and Data section presents the empirical methodology and data. The econometric results are presented and discussed in the Estimation and Results section, and concluding remarks follow in the last section.

Literature Review

We examine extant literature on the relationships between school infrastructure/teachers’ human capital and the academic performance of primary students. With each relationship in mind, we also focused on the research on the primary education system in Brazil.

School Infrastructure

A review by Glewwe et al. (2011) of the early literature (published between 1990 and 2010) investigates which specific school characteristics, if any, have positive impacts on education attainment. Based on the analyses of more than 75 high-quality studies, they conclude that almost all of the estimated impacts on time in school and learning of most school characteristics are statistically insignificant, especially when the evidence is limited to the highest-quality studies. The availability of desks is one of the few variables that has a positive effect on academic performance (Glewwe et al., 2011). More recent studies, such as Glewwe et al. (2020) and Hanushek (2020), further explore the role of school infrastructure on academic performance. The later studies point out that commonly purchased inputs in schools, including class size, teacher experience, and teachers’ education levels, exhibit weak relationships to student outcomes, suggesting that traditional education policies to improve input usage may not improve achievement (Hanushek, 2020). Hanushek (2020), however, reports that differences in teacher quality, defined in terms of effects on student performance, are important, even if they are not closely tied to salaries or other teacher attributes.

An early study by Branham (2004) employed data from 226 Independent School District schools in Houston in the US to explore the effect of school infrastructure on student attendance and dropout rates. Although the study found that the quality of school infrastructure has a direct effect on school attendance, it does not offer similar evidence regarding academic performance. That missing link is, however, provided in Durán-Narucki (2009), which examined data from 95 elementary schools in New York City on school building conditions and the results of standardized tests on English and math, and concluded that students attending schools with sub-standard facilities attended fewer days on average and had lower test scores on both subjects. Hong and Zimmer (2016) examined the causal impact of capital expenditures on school district proficiency rates in Michigan and found that these have positive effects on the academic proficiency levels of seventh grade students.

More recently, Adukia (2017) investigated the importance of adequate sanitation on the educational attainment of young girls in India using school-level data to show mixed results in that although school bathroom construction substantially increased the enrollment of young girls, academic scores did not increase following their construction. Conlin and Thompson (2017) analyzed the Ohio School Facilities Commission Classroom Facilities Assistance Program, which disbursed over $10 billion toward the improvement of local school facilities in 231 school districts between 1997 and 2011. They found, in line with Durán-Narucki (2009) and Hong and Zimmer (2016), that the percentage of students in the school district testing proficient in math and reading increased once the new and renovated buildings were ready (Conlin & Thompson, 2017). Lastly, Cuesta et al. (2016) examined the economics and education literature published from 1990 to 2012 to assess the extent to which specific types of school infrastructure have a causal impact on student learning and enrollment. They present evidence that school libraries and the creation of new schools lead to improved learning and enrollment. The literature also provides evidence that access to adequate bathroom facilities improves student learning, while laboratories and drinking water facilities increase enrollment (Cuesta et al., 2016).

Regarding research on this strand on Brazil, Franco et al. (2007) analyzed the relationship between school efficiency, as measured by the average performance of schools, and school characteristics associated with school performance. They found that infrastructure and school resources make a difference in education, but they vary widely across the country. Alves and Soares (2013) explored the relationship between IDEB scores and school infrastructure in Brazil and found, when controlling for other characteristics, that school infrastructure has a positive impact on IDEB scores. Pieri and Santos (2014) used factor analysis to explore the School Infrastructure Index (IIE), which captures student access to running and filtered water, sewage, concrete buildings, garbage collection, electricity, sports facilities, libraries, restrooms, laboratories, media equipment, computers, and printers.⁷ They found that IIE explains 23% of the variation in academic performance, and that academic equipment and physical plant variables have the greatest impact on proficiency exam results. Most recently, Matos and Rodrigues (2016) examined 36,236 schools to evaluate the main factors allowing public schools in Brazil to reach the IDEB target; they reveal that for elementary schools, the most important factor is school infrastructure, but school complexity exerts a negative impact on a school’s ability to maintain IDEB standards (Matos & Rodrigues, 2016). However, as pointed out earlier, Pieri and Santos (2014) and Matos and Rodrigues (2016) employ hierarchical and logistic approaches, respectively, which did not explain how key intra-school and socioeconomic factors influence schools with different IDEB results.

Teachers’ Human Capital

Friedman’s (1964) seminal study examined educational background and its relation to compensation of public school teachers in New York City when the city school board opted to hire teachers with post-baccalaureate education levels, beyond those required by teacher certification processes. Since this early study, attention has shifted to the importance of teachers’ human capital in boosting academic performance. In one such example, Croninger et al. (2007) assert that a primary function of education policy is to assess whether teacher qualifications such as certification status, degree level, preparation, and experience predict student achievement. This study, which employs data from the Early Childhood Longitudinal Study (ECLS) to analyze the relationship between elementary school teacher qualifications and first-grade achievement in reading and mathematics, has found positive effects of teachers’ degree type and experience on reading achievement, while certification status is not a significant determinant of academic performance (Croninger et al., 2007).

Metzler and Woessmann (2012) estimated the causal effect of teacher subject knowledge on student achievement using a unique Peruvian dataset that tested both sixth-grade students and their teachers in math and reading. They find that one standard deviation in subject-specific teacher achievement increases student achievement by 9% of the standard deviation in math. However, the effects in reading are found to be much smaller and, in most cases, not statistically significant (Metzler & Woessmann, 2012). Next, Cebula et al. (2015) used Academic Performance Index (API) data from high schools in Los Angeles County (California) to examine the relationships between school performance (at the high school level) on standardized exams and both teacher pay/salaries and teacher quality, where the latter is measured by teachers’ human capital. They show that both teacher pay and teacher quality exercise positive impacts on school performance—an increase from the lowest to the highest range of teacher pay and teacher quality generates academic performance improvement of approximately three and six percentage points, respectively (Cebula et al., 2015).

In terms of research using data from Brazil, Pieri and Santos (2014) used factor analysis to explore the Index of Instructor Education (IFP), which captures teachers’ human capital.⁸ They report that the IFP explains 45% of the variation in academic performance and that the most important variable is whether the teacher holds a bachelor’s degree. Interestingly, holding a “magistério” (teaching certificate) exerts a negative impact on the proficiency exams results. Finally, da Fonseca and Namen (2016) investigated how instructor profiles influence math instruction; they find that low wages impede math learning.

Methodology and Data

Prior empirical studies on academic performance have employed hierarchical linear models (Almeida et al., 2013) and logistic regression (Koslinski et al., 2014; Matos & Rodrigues, 2016). However, these methodologies fail to explain how prominent intra-school and socioeconomic factors influence schools with different IDEB results. We address these empirical shortcomings by employing quantile regression to explore the determinants of academic performance across schools in Brazil. This approach follows Koenker and Bassett (1978), who present a quantile regression model that yields robust estimators from the quantiles of the dependent variable, and Koenker and Machado (1999), who improved upon this foundational methodology. One of the advantages of quantile regression is that it allows for the analysis of the full range of the dependent variable while also adjusting the independent variables in each quantile, which addresses heteroskedasticity. We apply the Koenker and Machado (1999) methodology to a variation in quantiles from .02 to .98, with increments of .01, to estimate the following regression specification⁹

{I D E B 15}_{i} = β_{0} + β_{1} {I n f r a S a n i}_{i} + β_{2} {I n f r a F a m i l y}_{i} + β_{3} {I n f r a P e r s o n n e l}_{i} + β_{4} {I n f r a L a b s}_{i} + β_{5} {I n f r a S p o r t s}_{i} + β_{6} {I n f r a I n f o}_{i} + β_{7} {I n f r a M e d i a}_{i} + β_{8} {C o m p u t e r s}_{i} + β_{9} {T o t S t u d e n t s}_{i} + β_{10} {S t a f f S t u d}_{i} + β_{11} {G r a d e L e v e l s}_{i} + β_{12} {N S h i f t s}_{i} + β_{13} {C l a s s S i z e}_{i} + β_{14} {T e a c h D e g r e e}_{i} + β_{15} {S o c i o E c o n}_{i} + ε_{i},

(1)

where the dependent variable, IDEB15_i, captures the academic performance score, as measured by IDEB, of each school i in our dataset. Given that prior research has highlighted the important role of access to education-related infrastructure, we include (1) the regressor InfraSani_i, an index capturing access in school i to sanitation-based infrastructure—running water, sewage, and garbage collection.¹⁰

Other indices included in (1) are InfraFamily_i, InfraPersonnel_i, and InfraLabs_i, and they constitute indices related to (a) access to childcare facilities, kitchens, and cafeterias; (b) access to teachers’ lounge, principal’s office, and other areas related to support of school personnel; and (c) access to educational laboratories and auditoriums. Each of these infrastructure variables is expected to be positively related to IDEB15 in (1) above, as are InfraSports_i, InfraInfo_i, and InfraMedia_i, the indices related to (d) access to sports-related infrastructure, such as playing fields and gymnasiums, (e) access to information-related resources, such as libraries, computing facilities, and internet resources, and (f) access to multimedia-related infrastructure, such as copiers, printers, projectors, and televisions, respectively.

Also included on the right-hand side of (1) is Computers_i, representing the number of computers available to students and teachers in school i. Burney et al. (2013) assert that this resource is particularly useful to both teachers and students. Therefore, Computers_i is expected to be positively related to IDEB15_i. The variable attached to β₁₀, StaffStud_i is equal to the number of school staff (including teachers) per student enrolled in school i. It is included as a secondary index of school infrastructure (see Pieri & Santos, 2014). Each is expected to carry a positively signed coefficient in the empirical estimation of (1).

Additionally, several school complexity variables are included in equation (1). These are TotStudents_i, GradeLevels_i, and NShifts_i, representing the number of students enrolled in school i, the number of grade levels covered by school i, and the number of school shifts offered by school i, respectively.¹¹ Research on school size as an element of school complexity by Lee et al. (2004) supports the notion that school size negatively affects student achievement. A negative expectation is similarly attached to GradeLevels_i and NShifts_i. Also included in school complexity is ClassSize_i, which is the average number of students per classroom at school i. Although Matos and Rodrigues (2016) do not find a statistically significant effect on academic performance in Brazil from this variable, Crahay (2007) and Pieri and Santos (2014) provided support for us to include this variable in (1).

The penultimate regressor in (1) is TeachDegree_i, which is the percentage of teachers employed by school i holding a college degree. In Brazil, the requisite qualification to teach in elementary schools is “magistério,” comparable to a high school diploma in the U.S. One of the kay targets of Brazil’s national plan (P. Brasil, 2014) is to raise this to a college degree. The literature emphasizes the importance of teachers’ human capital in shaping the academic performance of primary students (e.g., Croninger et al., 2007; Friedman, 1964; Metzler & Woessmann, 2012). Cebula et al. (2015) have found that teacher qualifications (human capital) support the academic performance of students in the U.S.; their findings are supported by Matos and Rodrigues for Brazil (2016). Thus, the parameter estimate attached to TeachDegree_i in (1) is expected to be positive.

Finally, the variable SocioEcon_i was included to capture the socioeconomic level of the parents of Brazil’s primary students. Specifically, it is the family socioeconomic level of the National Literacy Assessment (ANA), organized into seven levels; level one being the lowest and level seven the highest. It is a latent variable inferred from the other observable variables and synthesizes information on parents’ education and household income (M. Brasil, 2014).¹² Several studies have examined the impact of socioeconomic status on academic achievement, including Cebula et al. (2015), which, in its examination of the academic performance index (API) across Los Angeles County high schools, includes free lunch recipients. A contemporaneous study by Firpo et al. (2015) for Brazil includes socioeconomic status in an analysis of the determinants of students’ outcomes in the fifth grade of elementary school. Based on these studies, we expect that SocioEcon_i will be positively related to IDEB15_i.

Each of the variables used in this study are listed and described in Table 1. The IDEB data were obtained from the 2015 school census¹³ with 35,560 observations in total. All independent variables are in standard form (i.e., a mean of zero and a standard deviation equal to one). All outliers (38 schools or 0% of the census) were withdrawn from the dataset. Thus, we analyzed data from 35,490 schools, including 5105 rural schools. Table 2 presents the means and standard deviations (by quantile) for both urban and rural schools. It shows that IDEP15 increases across quartiles, ranging from about 4.0 to 6.7 for urban schools, and from about 3.2 to 6.7 for rural schools. Urban schools also have greater infrastructure across all categories than their rural counterparts. Compounding the relative deficiencies facing rural schools is the fact that rural school complexity generally exceeds that of their urban counterparts, while both teachers’ human capital and the socioeconomic level of school parents are greater in urban than in rural environments.

Table 1.

Variable Names and Descriptions.

Variable	Description
IDEB15 _i	The academic performance score, as measured by IDEB, for each school i
InfraSani _i	An index capturing access in school i to sanitation-related infrastructure, such as access to running water, sewage, and garbage collection
InfraFamily _i	An index capturing access in school i to family related infrastructure, such as access to childcare facilities, kitchens, and cafeterias
InfraPersonnel _i	An index capturing access in school i to personnel-related infrastructure, such as access to teachers’ lounge, principal’s office, and other areas related to support of school personnel
InfraLabSup _i	An index capturing access in school i to infrastructure such as educational laboratories and auditoriums
InfraSports _i	An index capturing access in school i to sports-related infrastructure, such as playing fields and gymnasiums
InfraInfo _i	An index capturing access in school i to information-related infrastructure, such as libraries, computing facilities, and internet resources
InfraMedia _i	An index capturing access in school i to multimedia-related infrastructure, such as copiers, printers, projectors, and televisions
Computers _i	The number of computers available to students and teachers in school i
TotStudents _i	The number of students enrolled in school i
StaffStud _i	The number of school staff (including teachers) per student in school i
GradeLevels _i	The number of grade levels covered by school i
NShifts _i	The number of school shifts offered by school i
ClassSize _i	The average number of students per classroom at school i
TeachDegree _i	The percentage of teachers employed by school i holding a college degree
SocioEcon _i	An index capturing the socioeconomic characteristics (e.g., education, income) pertaining to the parents of the students enrolled in school i

Table 2.

Summary Statistics by Quartile.

Variables		Bottom Quartile		Second Quartile		Third Quartile		Top Quartile
Variables		Urban Schools	Rural Schools	Urban Schools	Rural Schools	Urban Schools	Rural Schools	Urban Schools	Rural Schools
IDEB15	(M)	4.040	3.226	5.111	4.016	5.836	4.764	6.715	6.015
IDEB15	(SD)	.480	.395	.241	.187	.198	.268	.439	.631
InfraSani	(M)	.995	.829	.997	.893	.998	.924	.998	.941
InfraSani	(SD)	.038	.184	.026	.145	.020	.131	.020	.119
InfraFamily	(M)	.438	.309	.510	.352	.539	.409	.546	.452
InfraFamily	(SD)	.200	.150	.197	.170	.193	.190	.189	.192
InfraPersonnel	(M)	.777	.562	.868	.634	.920	.706	.939	.759
InfraPersonnel	(SD)	.273	.304	.226	.303	.179	.298	.155	.280
InfraLabSup	(M)	.098	.030	.125	.034	.134	.054	.141	.061
InfraLabSup	(SD)	.229	.122	.259	.134	.269	.176	.281	.183
InfraSports	(M)	.330	.203	.438	.251	.524	.337	.583	.433
InfraSports	(SD)	.231	.184	.253	.204	.244	.244	.239	.259
InfraInfo	(M)	.715	.352	.883	.454	.878	.560	.900	.659
InfraInfo	(SD)	.293	0.312	.228	.322	.192	.316	.171	.287
InfraMedia	(M)	12.03	7.725	15.94	9.684	18.39	11.89	20.68	13.68
InfraMedia	(SD)	6.875	4.784	9.115	5.458	9.770	6.440	10.80	7.152
Computers	(M)	14.24	5.748	18.40	7.539	21.60	9.664	23.82	12.70
Computers	(SD)	13.22	6.104	13.12	9.814	15.35	7.879	19.87	17.54
TotStudents	(M)	591.6	398.9	607.4	416.7	618.3	404.3	589.0	351.9
TotStudents	(SD)	369.0	223.6	357.3	248.7	368.2	243.5	339.9	207.3
StaffStud	(M)	6.074	5.036	6.277	4.986	6.548	5.139	7.060	5.079
StaffStud	(SD)	3.281	2.615	3.500	2.687	3.512	2.838	3.715	2.654
GradeLevels	(M)	2.446	3.013	2.379	2.908	2.263	2.756	2.089	2.573
GradeLevels	(SD)	.950	.926	.956	.925	.933	.911	.946	.868
NShifts	(M)	2.439	2.442	2.342	2.436	2.299	2.288	2.245	2.160
NShifts	(SD)	.535	.551	.516	.554	.505	.531	.497	.507
ClassSize	(M)	23.73	21.67	23.57	21.64	23.74	21.49	24.06	20.74
ClassSize	(SD)	4.057	3.808	4.034	3.985	4.056	4.034	4.007	3.759
TeachDegree	(M)	.676	.446	.797	.533	.856	.641	.884	.746
TeachDegree	(SD)	.270	.331	.216	.324	.167	.304	.137	.253
SocioEcon	(M)	3.524	2.238	4.266	2.664	4.744	3.235	5.175	3.727
SocioEcon	(SD)	.842	.763	.849	.825	.759	1.028	.794	1.133

Note. (M) = mean, (SD) = standard deviation.

The correlation coefficient estimates for the variables included in (1) are provided in Table 3. The top part presents the correlations for the variables based on the data for urban schools. Only two correlation coefficients out of the 106 not involving the dependent variable, IDEB15, signify a high degree of correlation (i.e., between ±.5 and ±1.0). These are −.55 for StaffStud and GradeLevels, and +.61 for NShifts and GradeLevels. Another 18 of these 106 correlation coefficients signify a moderate degree of correlation (i.e., between ±.3 and ±.5). Correlations from the rural school data are shown in the bottom part, where only one correlation coefficient out of the 106 not involving the dependent variable, IDEB15, signifies a high degree of correlation (i.e., between ±.5 and ±1.0). NShifts and GradeLevels show the coefficient, +.59. In the case of rural schools, 25 of these 106 correlation coefficients signify a moderate degree of correlation (i.e., between ±.3 and ±.5). Lastly, none of the 28 coefficients involving SocioEcon, though not the dependent variable IDEB15, indicate a high degree of correlation, whereas only 12 of these 28 coefficients exhibit a moderate degree of correlation.¹⁴

Table 3.

Correlation Coefficients.

Urban Schools	GradeLevels	IDEB15	TotStudents	ClassSize	StaffStud	TeachDegree	NShifts	SocioEcon	InfraSani	InfraFamily	InfraPersonnel	InfraMedia	InfraLabSup	InfraSports	InfraInfo
Computers	.14	.22	.25	.04	−.11	.15	.10	.24	.04	.22	.21	.31	.20	.24	.33
GradeLevels		−.15	.43	−.07	−.55	−.03	.61	−.07	.00	.11	.13	.10	.26	.11	.13
IDEB15			.00	.03	.11	.36	−.14	.61	.05	.21	.27	.33	.07	.36	.29
TotStudents				.41	−.11	.05	.42	.08	.03	.20	.25	.21	.28	.15	.19
ClassSize					.32	.01	.02	.07	.01	.01	.07	.10	−.03	.01	.00
StaffStud						.02	−.27	.08	.00	−.06	−.09	−.01	−.23	−.05	−.15
TeachDegree							−.02	.31	.05	.15	.24	.22	.02	.20	.23
NShifts								−.11	−.01	.05	.05	.02	.13	−.02	.06
SocioEcon									.10	.29	.36	.35	.16	.43	.34
InfraSani										.06	.07	.05	.02	.07	.07
InfraFamily											.38	.28	.25	.38	.33
InfraPersonnel												.26	.18	.39	.39
InfraMedia													.18	.34	.32
InfraLabSup														.23	.20
InfraSports															.34

Rural Schools	GradeLevels	IDEB15	TotStudents	ClassSize	StaffStud	TeachDegree	NShifts	SocioEcon	InfraSani	InfraFamily	InfraPersonnel	InfraMedia	InfraLabSup	InfraSports	InfraInfo
Computers	.06	.22	.16	.02	−.08	.14	.02	.24	.15	.25	.25	.33	.12	.27	.36
GradeLevels		−.18	.39	−.05	−.48	−.11	.59	−.22	−.15	.00	.10	.03	.07	−.01	.06
IDEB15			−.08	−.08	.00	.33	−.20	.50	.26	.28	.23	.34	.07	.35	.34
TotStudents				.32	−.06	−.01	.41	.00	.10	.16	.26	.20	.13	.08	.19
ClassSize					.30	.00	.04	.03	.08	.05	.05	.05	.00	−.01	.02
StaffStud						.04	−.20	.11	.09	−.02	−.13	−.06	−.08	−.05	−.14
TeachDegree							−.11	.37	.25	.20	.23	.25	.07	.25	.24
NShifts								−.17	−.03	−.04	.00	−.02	.03	−.10	.00
SocioEcon									.42	.37	.35	.37	.15	.44	.37
InfraSani										.26	.27	.30	.05	.25	.32
InfraFamily											.43	.39	.21	.44	.41
InfraPersonnel												.37	.17	.43	.43
InfraMedia													.14	.40	.41
InfraLabSup														.22	.17
InfraSports															.39

Estimation and Results

Given that the independent variables included in equation (1) are normalized, the regression coefficients are the indicators of the relative impact of a given variable on IDEB15. We estimate separate OLS regressions for urban and rural schools. In total, data from 30,385 urban schools were analyzed. Table 4 presents the estimated coefficients from an OLS model including the education variables and the socioeconomic level. Both R² and the adjusted R² for the model were .440. Most infrastructure-related variables are positively and significantly related to IDEB15 for urban schools. Thus, greater access to sanitation, personnel, sports, information, and multi-media infrastructure is positively associated with students’ academic performance in urban schools. Moreover, the results in Table 4 suggest that access to computers is associated with a higher level of academic performance.

Table 4.

OLS Regression Results for Urban Schools: Dependent Variable = IDEB15.

Variable	Coefficient Estimate	t-ratio	Sig. Level
Intercept	5.275	1014.8	***
InfraSani	.051	4.246	***
InfraFamily	−.020	−3.85	***
InfraPersonnel	.020	3.362	***
InfraLabSup	−.022	−4.64	***
InfraSports	.094	16.99	***
InfraInfo	.066	10.53	***
InfraMedia	.097	19.06	***
Computers	.060	11.96	***
TotStudents	−.044	−7.18	***
StaffStud	.030	5.181	***
GradeLevels	−.108	−15.47	***
NShifts	−.008	−1.41
ClassSize	−.009	−1.66	*
TeachDegree	.188	34.18	***
SocioEcon	.535	85.58	***
	n = 30,385 R² = .440 adj. R² = .440

Note. ***[*] denotes significance at the .01[.10] level.

Interestingly, larger schools and those offering more study shifts enroll students whose academic performance trails that of smaller schools and schools offering fewer study shifts. However, the result for the latter groups is not significant. Urban schools encompassing many grade levels and larger class sizes also enroll students whose academic performance lags behind that of schools with fewer complexities. A larger school staff relative to the student body is associated, as expected, with greater academic performance. Lastly, teachers’ human capital, proxied by TeachDegree, is positively and significantly related to IDEB15. This is a noteworthy result that supports those of Croninger et al. (2007), Metzler and Woessmann (2012), and Cebula et al. (2015).

A quantile regression-based graphical analysis of urban schools is provided in Figure 1. We present the estimator (on the y-axis) of each variable with respect to the quantile intervals examined (on the x-axis), where the red lines represent the coefficients from the OLS regression, and the black dots and gray areas represent the coefficients from the quantile regression and confidence intervals, respectively. Schools with the lowest IDEB15 scores reside in quantiles close to zero in each diagram, while each diagram also includes the average value of the coefficient along with its standard errors. Where the value on the y-axis is positive, the variable is positively related to IDEB15.

Figure 1.

Quantile regression for urban schools: Dependent variable = IDEB15.

The results in Figure 1 indicate that InfraInfo and Computers were both positively related to variations in IDEB15 across urban schools in Brazil. Additionally, TeachDegree is positively related to variations in IDEB15 across urban schools in Brazil. This result supports those of Croninger et al. (2007), Metzler and Woessmann (2012), and Cebula et al. (2015), and may also be linked to differences across schools regarding teaching methods and other factors (see Haguette et al., 2016). Interestingly, the correlation between socioeconomic status and academic performance across schools is weaker among higher-achieving schools. Li et al. (2020), citing Costello et al. (2003) and Neville et al. (2013), point out that improving family economic conditions reduces children’s risk of psychiatric disorders, and that interventions targeting individual attributes (e.g., attention) can significantly facilitate child development in low-socioeconomic-status families. Given that relatively less access to therapies for attention deficits and other issues that may hinder academic performance increases with a rise in socioeconomic status, the stronger correlation between academic achievement and socioeconomic status across lower- and mid-achieving schools is less surprising. Li et al.’s research (2020) on how socioeconomic conditions and individual characteristics or attributes, such as self-efficacy, are interdependent forces that influence each other may also play a role in this correlation asymmetry.

Next, the data from 5105 rural schools were analyzed in the same way as urban schools. Table 5 presents the coefficients from OLS estimation, including the intra-school variables and socioeconomic level. The R² for the model was .343, while the adjusted R² for the model was .341. Table 5 reveals that most infrastructure-related variables are positively and significantly related to IDEB15 for rural schools, that is, access to family, sports, information, and multi-media infrastructure is positively associated with the academic performance of students. Additionally, access to computers is associated with a higher level of academic performance.

Table 5.

OLS Regression Results for Rural Schools: Dependent Variable = IDEB15.

Variable	Coefficient Estimate	t-ratio	Sig. Level
Intercept	5.179	215.3	***
InfraSani	.004	.588
InfraFamily	.033	1.97	**
InfraPersonnel	−.025	−1.98	**
InfraLabSup	−.049	−2.35	**
InfraSports	.104	6.066	***
InfraInfo	.113	8.515	***
InfraMedia	.185	8.105	***
Computers	.061	3.165	***
TotStudents	−.088	−3.64	***
StaffStud	−.049	−2.48	**
GradeLevels	−.079	−4.04	***
NShifts	−.070	−4.54	***
ClassSize	−.081	−5.47	***
TeachDegree	.107	10.12	***
SocioEcon	.339	19.91	***
	n = 5105 R² = .343 adj. R² = .341

Note. ***(**) denotes significance at the .01(.05) level.

As before, larger schools and those offering a greater number of study shifts and larger average class sizes enroll students whose academic performance trails that of schools that are smaller and offering fewer shifts and larger average class sizes. Rural schools encompassing a large number of grade levels also enroll students whose academic performance lags behind that of their less complex counterparts. Another area where rural schools differ from their urban counterparts is StaffStud: this variable was unexpectedly negatively and significantly related to IDEB15. Teachers’ human capital, proxied by TeachDegree, is positively and significantly related to students’ academic performance in the case of rural schools, an important finding that supports Croninger et al. (2007), Metzler and Woessmann (2012), and Cebula et al. (2015).

A quantile regression-based graphical analysis for rural schools is provided in Figure 2. The results indicate that InfraInfo and InfraMedia are both positively related to variations in IDEB15 across rural schools in Brazil. Additionally, TeachDegree is positively related to variations in IDEB15 across rural schools in Brazil. This result supports those of Croninger et al. (2007), Metzler and Woessmann (2012), and Cebula et al. (2015), and may also be linked to differences across schools regarding teaching methods and other factors (see Haguette et al., 2016). As in the case of urban schools, the correlation between socioeconomic status and academic performance is weaker among higher-achieving schools. Again, Li et al.’s (2020) research citing Costello et al. (2003) and Neville et al. (2013) on the relationships between improvements in the socioeconomic status of families and (1) access to mental health therapies and (2) students’ concepts of identity and self, offer at least a partial explanation for this correlation asymmetry.

Figure 2.

Quantile regression for rural schools: Dependent variable = IDEB15.

Table 6 summarizes the main quantile regression results for both urban and rural schools (dependent variable = IDEB15). Based on Figures 1 and 2 and Table 6, most quantile patterns are similar between rural and urban schools, with some exceptions. These pertain to InfraLabSup, InfraMedia, and TotStudents, which have different quantile results for the rural and urban samples. These differences may be related to the finding that rural school complexity generally exceeds urban school complexity. As indicated in Table 2, the average number of grade levels and school shifts covered by rural schools exceeded that of their urban counterparts in six of the eight comparable quartiles. Additionally, Wang (2013) indicates that there are significant differences in the attitudes, competence levels, and experience with technology integration between rural and urban elementary school teachers. Adding to this concern is the lower access to media technologies in the homes of rural students. Latest research by Hampton et al. (2021) finds that rural students with broadband home internet access show more interest in class and are able to participate in a more diverse array of online media activities than rural students lacking such access usually do. Given that rural areas in the U.S. experience some of the largest inequalities in broadband internet access (Hampton et al., 2021), this is a vexing issue in Brazil. Lastly, of the 22 results summarized in Table 6, 19 relate to quantile regression results that support prior literature, that is, 86.4% of the results generated by this study support their counterparts in earlier studies. Additionally, only 8.1% of results are in conflict with those of prior studies.

Table 6.

Summary of Main Quantile Regression Results for Urban and Rural Schools: Dependent Variable = IDEB15.

Variable	Environment	Degree of Impact	Prior Literature	Finding Relative to Prior literature	More influenced IDEB15
InfraLabSup	Urban	Very low	High is best	Conflicts with prior literature	Lower IDEB
InfraLabSup	Rural	Very low	High is best	Conflicts with prior literature	Higher IDEB
InfraSports	Urban	Low	High is best	Supports prior literature	Lower IDEB
InfraSports	Rural	Middle	High is best	Supports prior literature	Lower IDEB
InfraInfo	Urban	Low	High is best	Supports prior literature	No difference
InfraInfo	Rural	Middle	High is best	Supports prior literature	Higher IDEB
InfraMedia	Urban	Low	High is best	Supports prior literature	Lower IDEB
InfraMedia	Rural	Middle	High is best	Supports prior literature	Higher IDEB
Computers	Urban	Low	High is best	Supports prior literature	Higher IDEB
Computers	Rural	Low	High is best	Supports prior literature	No difference
TotStudents	Urban	Very low	Low is best	Supports prior literature	Higher IDEB
TotStudents	Rural	Low	Low is best	Supports prior literature	Higher IDEB
GradeLevels ^a	Urban	Middle	Low is best	Supports prior literature	Lower IDEB
GradeLevels	Rural	Low	Low is best	Supports prior literature	Higher IDEB
NShifts ^a	Urban	Very low	Low is best	Supports prior literature	No difference
NShifts	Rural	Low	Low is best	Supports prior literature	Higher IDEB
ClassSize	Urban	Very low	Low is best	No conclusion	No difference
ClassSize	Rural	Low	Low is best	Supports prior literature	Higher IDEB
TeachDegree	Urban	Middle	High is best	Supports prior literature	Higher IDEB
TeachDegree	Rural	Middle	High is best	Supports prior literature	Higher IDEB
SocioEcon	Urban	Very high	High is best	Supports prior literature	Lower IDEB
SocioEcon	Rural	Very high	High is best	Supports prior literature	No difference^b

^aNo significance level in OLS.

^bFor most schools, there was no difference. However, for schools with a high IDEB (quantile >.8), the impact was smaller.

Conclusions, Implications, and Limitations

This study employs quantile regression to examine relationships between the academic performance of primary school students in Brazil and various demographic and institutional variables, such as socioeconomic characteristics, variations in school infrastructure and school complexity, and teachers’ human capital thereby extending the education economics literature by addressing the empirical shortcomings of prior research.

General Conclusions

Our results based on data from 35,490 primary schools in Brazil suggest that several institutional variables in our model specification, including access to school libraries, computer facilities, projectors, and televisions, are positively and significantly related to the academic performance of primary students in Brazil’s public education system. We also found that parents’ socioeconomic level is positively associated with students’ academic performance. However, school complexity, captured in the number of grade levels and school shifts at primary schools, is negatively related to academic performance.

Theoretical and Practical Implications

The results related to parents’ socioeconomic level represent a contextual condition with a high impact on school results, affecting urban schools on a larger scale. In rural schools, its impact is somewhat milder, while in both cases, high-performing schools are less affected by this condition. The implication that low socioeconomic conditions are highly correlated with low incomes demonstrates that structural issues in society affect school performance, at least according to the standards currently used for their assessment. This implication supports studies conducted since the publication of Equality of Educational Opportunity (Coleman et al., 1966) and subsequent suggestions by Alves and Soares (2013) that the results of research on academic performance must be disclosed within a context, given that they must be seen as an effective expression of the social function of student learning in the environment they inhabit. Therefore, academic performance should not be viewed solely as the responsibility of school instructors and employees.

Finally, the negative relationship between school complexity and academic performance manifests differently across urban and rural schools. For urban schools, this relationship is stronger in schools with low IDEB, while in rural schools, this relationship is reversed. This result is reinforced by the negative relationship between the total number of students and school performance, which is more pronounced for schools with high IDEB. Thus, structural issues concerning the use of facilities and management of school units must be considered in the formulation of policies. In other words, decentralization of school levels into smaller units can contribute to better results. However, this is a complex issue, especially in view of the availability of resources and logistical infrastructure, which is mostly poor in developing countries, and innovative solutions, such as the separation of academic units within the same school facility, are needed.

Limitations and Future Directions

Despite our study extending the literature, several avenues remain for future research on school infrastructure and the academic performance of primary school students in Brazil. For example, future research could attempt to capture omitted variables that may increase the precision of our statistical findings and allow for policy recommendations regarding academic performance. The identification of factors that improve the academic performance of schools at the bottom of the socioeconomic scale would better inform public policy. Additionally, given that the qualification of teachers, as represented by the percentage of teachers with higher education, is the intra-school variable that most impacts IDEB (in both urban and rural schools), future research along the lines of Cebula et al. (2015) might shed more light on the benefits of additional investments in teacher qualifications. Finally, endogeneity was not formally addressed in our study, and future research might also examine a system of equations wherein some of the regressors examined here are jointly determined by the academic performance of schools. Such an analysis may, for example, determine whether high-achieving students go into schools with better infrastructure. A positive finding in such a case would suggest that higher academic achievement is the result of sorting, and not better infrastructure.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Franklin G. Mixon

Notes

References

Adukia

(2017). Sanitation and education. American Economic Journal: Applied Economics, 9(2), 23–59. https://doi.org/10.1257/app.20150083

Almeida

L. C.

Dalben

Freitas

L. C. D.

(2013). O Ideb: Limites e ilusões de uma política educacional. Educação & Sociedade, 34(1), 1153–1174. https://doi.org/10.1590/s0101-73302013000400008

Alves

M. T. G.

Soares

J. F.

(2013). Contexto escolar e indicadores educacionais: Condições desiguais para a efetivação de uma política de avaliação educacional. Educação e Pesquisa, 39(1), 177–194. https://doi.org/10.1590/s1517-97022013000100012

Barro

R. J.

(1997). Getting it right: Markets and choices in a free society. The MIT Press.

Barro

R. J.

(1998). Determinants of economic growth: A cross-country empirical study. The MIT Press.

Branham

(2004). The wise man builds his house upon the rock: The effects of inadequate school building infrastructure on student attendance. Social Science Quarterly, 85(5), 1112–1128. https://doi.org/10.1111/j.0038-4941.2004.00266.x

Brasil

M. E. C.

(2014a). Planejando a Próxima Década Conhecendo as 20 Metas do Plano Nacional de Educação. MEC.

Brasil

(2014b). Aprova o Plano Nacional de Educação e dá outras providências. Planalto.

Burney

N. A.

Johnes

Al-Enezi

Al-Musallam

(2013). The efficiency of public schools: The case of Kuwait. Education Economics, 21(4), 360–379. https://doi.org/10.1080/09645292.2011.595580

10.

Carnoy

Marotta

Louzano

Khavenson

Guimarāes

F. R. F.

Carnauba

(2017). Intranational comparative education: What state differences in student achievement can teach us about improving education – the case of Brazil. Comparative Education Review, 61(4), 726–759. https://doi.org/10.1086/693981

11.

Casassus

(2009). Uma nota crítica sobre a avaliação estandardizada: A perda de qualidade e a segmentação social. Sísifo: Revista de Ciências da Educação, 9(1), 71–78.

12.

Cebula

R. J.

Mixon

F. G.

Jr. Montez

(2015). Teachers’ salaries and human capital, and their effects on academic performance: An institution-level analysis of Los Angeles County high schools. Journal of Economics and Finance, 39(2), 347–356. https://doi.org/10.1007/s12197-013-9261-3

13.

Coleman

J. S.

Campbell

E. Q.

Hobson

C. J.

McPartland

Mood

A. M.

Weinfeld

F. D.

York

R. L.

(1966). Equality of educational Opportunity. U.S. Department of Health, Education, and Welfare.

14.

Conlin

Thompson

P. N.

(2017). Impacts of new school facility construction: An analysis of a state-financed capital subsidy program in Ohio. Economics of Education Review, 59(1), 13–28. https://doi.org/10.1016/j.econedurev.2017.05.002

15.

Costello

E. J.

Compton

S. N.

Keeler

Angold

(2003). Relationships between poverty and psychopathology: A natural experiment. JAMA, 290(15), 2023–2029. https://doi.org/10.1001/jama.290.15.2023

16.

Crahay

(2007). Qual pedagogia para aos alunos em dificuldade escolar? Cadernos de Pesquisa, 37(130), 181–208. https://doi.org/10.1590/s0100-15742007000100009

17.

Croninger

R. G.

Rice

J. K.

Rathbun

Nishio

(2007). Teacher qualifications and early learning: Effects of certification, degree, and experience on first-grade student achievement. Economics of Education Review, 26(3), 312–324. https://doi.org/10.1016/j.econedurev.2005.05.008

18.

Cuesta

Glewwe

Krause

(2016). School infrastructure and educational outcomes: A literature review, with special reference to Latin America. Economia, 17(1), 95–130. https://doi.org/10.1353/eco.2016.a634033

19.

da Fonseca

S. O.

Namen

A. A.

(2016). Mineração em Bases de Dados do INEP: Uma análise exploratória para nortear melhorias no sistema educacional Brasileiro. Educação em Revista, 32(1), 133–157. https://doi.org/10.1590/0102-4698140742

20.

Duarte

N. D. S.

(2013). O impacto da pobreza no Ideb: um estudo multinível. Revista Brasileira de Estudos Pedagógicos, 94(237), 343–363. https://doi.org/10.24109/2176-6681.rbep.94i237.369

21.

Durán-Narucki

(2009). School building condition, school attendance, and academic achievement in New York City public schools: A mediation model. Journal of Environmental Psychology, 28(3), 278–286. https://doi.org/10.1016/j.jenvp.2008.02.008

22.

Firpo

Jales

Pinto

(2015). Measuring peer effects in the Brazilian school system. Applied Economics, 47(32), 3414–3438. https://doi.org/10.1080/00036846.2015.1016207

23.

Franco

Ortigão

Albernaz

Â.

Bonamino

Aguiar

Alves

Sátyro

(2007). Qualidade e eqüidade em educação: Reconsiderando o significado de “fatores intra-escolares. Ensaio: Avaliação e Políticas Públicas em Educação, 15(55), 277–298. https://doi.org/10.1590/s0104-40362007000200007

24.

Freitas

L. C.

de Sordi

M. R. L.

Malavasi

M. M. S.

de Freitas

H. C. L.

(2017). Avaliação Educacional: Caminhando Pela Contramão. Editora Vozes Limitada.

25.

Friedman

C. H.

(1964). Education of New York City public school teachers: An economic analysis. Industrial and Labor Relations Review, 18(1), 20–31. https://doi.org/10.2307/2520521

26.

Glewwe

Lambert

Chen

(2020). Education production functions: Updated evidence from developing countries. In Bradley

Green

(Eds.), The economics of education (pp. 183–215). Academic Press.

27.

Glewwe

P. W.

Hanushek

E. A.

Humpage

S. D.

Ravina

(2011). School resources and educational outcomes in developing countries: A review of the literature from 1990 to 2010. National Bureau of Economic Research. Working Paper 17554.

28.

Haguette

Pessoa

M. K. M.

Vidal

E. M.

(2016). Dez escolas, dois padrões de qualidade. Uma pesquisa em dez escolas públicas de Ensino Médio do Estado do Ceará. Ensaio: Avaliação e Políticas Públicas em Educação, 24(92), 609–636. https://doi.org/10.1590/s0104-40362016000300005

29.

Hampton

K. N.

Robertson

C. T.

Fernandez

Shin

Bauer

J. M.

(2021). How variation in internet access, digital skills, and media use are related to rural student outcomes: GPA, SAT, and educational aspirations. Telematics and Informatics, 63(1), 101666. https://doi.org/10.1016/j.tele.2021.101666

30.

Hanushek

E. A.

(2011). The economic value of higher teacher quality. Economics of Education Review, 30(3), 466–479. https://doi.org/10.1016/j.econedurev.2010.12.006

31.

Hanushek

E. A.

(2020). Education production functions. In Bradley

Green

(Eds.), The economics of education (pp. 161–170). Academic Press.

32.

Hong

Zimmer

(2016). Does investing in school capital infrastructure improve student achievement? Economics of Education Review, 53(1), 143–158. https://doi.org/10.1016/j.econedurev.2016.05.007

33.

INEP . (2013). Nota técnica - Indicador de Nível Socioeconômico das Escolas de Educação Básica (INSE) participantes da Avaliação Nacional da Alfabetização (ANA). INEP.

34.

INEP . (2014). Indicador para mensurar a complexidade da gestão na s escolas a partir dos dados do Censo Escolar da Educação Básica. INEP.

35.

Koenker

Bassett

Jr. (1978). Regression quantiles. Econometrica, 46(1), 33–50. https://doi.org/10.2307/1913643

36.

Koenker

Machado

J. A. F.

(1999). Goodness of fit and related inference processes for quantile regression. Journal of the American Statistical Association, 94(448), 1296–1310. https://doi.org/10.1080/01621459.1999.10473882

37.

Koslinski

M. C.

Cunha

C. P. D.

Andrade

F. M. D.

(2014). Accountability escolar: Um estudo exploratório do perfil das escolas premiadas. Estudos em Avaliação Educacional, 25(59), 108–137. https://doi.org/10.18222/eae255920142865

38.

Lee

V. E.

Franco

Abernaz

(2004). Quality and equality in Brazilian secondary schools: A multilevel cross-national school effects study. Unpublished Paper.

39.

Xia

(2020). Relationship between SES and academic achievement of junior high school students in China: The mediating effect of self-concept. Frontiers in Psychology, 10(1), 2513. https://doi.org/10.3389/fpsyg.2019.02513

40.

Matos

D. A. S.

Rodrigues

E. C.

(2016). Indicadores educacionais e contexto escolar: Uma análise das metas do Ideb. Estudos em Avaliação Educacional, 27(66), 662–688. https://doi.org/10.18222/eae.v27i66.4012

41.

Mbiti

I. M.

(2016). The need for accountability in education in developing countries. Journal of Economic Perspectives, 30(3), 109–132. https://doi.org/10.1257/jep.30.3.109

42.

Metzler

Woessmann

(2012). The impact of teacher subject knowledge on student achievement: Evidence from within-teacher within-student variation. Journal of Development Economics, 99(2), 486–496. https://doi.org/10.1016/j.jdeveco.2012.06.002

43.

Neto

J. L. H.

(2010). Avaliação externa de escolas e sistemas: Questões presentes no debate sobre o tema. Revista Brasileira de Estudos Pedagógicos, 91(227), 84–104.

44.

Neto

J. L. H.

Junqueira

R. D.

de Oliveira

A. S.

(2016). Do Saeb ao Sinaeb: Prolongamentos críticos da avaliação da educação básica. Em Aberto, 29(96), 21–37.

45.

Neville

H. J.

Stevens

Pakulak

Bell

T. A.

Fanning

Klein

Isbell

(2013). Family-based training program improves brain function, cognition, and behavior in lower socioeconomic status preschoolers. Proceedings of the National Academy of Sciences of the United States of America, 110(29), 12138–12143. https://doi.org/10.1073/pnas.1304437110

46.

OECD . (2017). Education at a glance 2017: OECD indicators. OECD Publishing.

47.

Pieri

R. G.

Santos

A. A.

(2014). Uma Proposta para o Índice de Infraestrutura Escolar e o Índice de Formação de Professores. INEP.

48.

Vidal

E. M.

Vieira

S. L.

(2011). Gestão educacional e resultados no Ideb: Um estudo de caso em dez municípios cearenses. Estudos em Avaliação Educacional, 22(50), 419–434. https://doi.org/10.18222/eae225020111963

49.

Wang

P.-Y.

(2013). Examining the digital divide between rural and urban schools: Technology availability, teachers’ integration level and students’ perception. Journal of Curriculum and Teaching, 2(2), 127–139. https://doi.org/10.5430/jct.v2n2p127