Introduction

There is broad political and social consensus in Brazil regarding the importance of public basic education²⁵. The repeated references to it in the Brazilian Constitution, as well as the requirement that a given percentage of each federative agent’s budget should go towards education, established by art. 212 of the Constitution²⁶, are proof of that. The political decision to constitutionally protect public basic education was the result of a mobilization of civil society during the drafting of the 1988 Constitution, the likes of which leaves no doubt about its centrality to Brazilian public life (^{Martins, 2018}).

Recent civil society initiatives, such as the movements: All for Education²⁷ and the Education Map²⁸, show that the commitment of Brazilians to the Basic Education agenda has not waned over the 30 years that separate us from the constitutional convention. This commitment continued to be strengthened in an institutional manner by the edition of the National Education Guidelines and Bases Law - LDB 9394/96, the National Curricular Guidelines - DCNs, 2013, the National Education Plans of 2010 and 2014 - PNEs, among other documents and regulations of national scope.

The specific choices made by representatives responsible for the 1988 Constitution are also informative with respect to another important aspect of our argument. They are symptomatic of the kind of attention that the law, through its regulatory instruments, has given to education: there is a focus on top-down²⁹ instruments, such as funding programs and federal resource transfers (such as for school meals, school transportation and textbooks), the definition of a national curriculum, teacher training programs, among others. To a large extent, this focus was an effective strategy to improve the initial phase of Basic Education - the initial years of Elementary School - in a significant way: since the beginning of the measurement of IDEB³⁰, there has been significant progress in the performance of Brazilian schools in the initial years of Elementary School³¹, demonstrating that public policies in the valorization and measurement of basic education (^{Coelho, 2008}) were, at least in part, successful.

This result makes sense, as the specialized literature indicates that factors such as school infrastructure (^{Alves; Soares, 2013b}), the socioeconomic level of students (Alves; Soares, 2007a), among other social and demographic variables (^{Andrade; Laros, 2007}), exert a significant influence on IDEB scores. In any case, much remains to be done: despite the adoption of the policies mentioned in the last PISA³², the country’s performance fell short of what was desirable (for an analysis of the poor results in the previous evaluation, see ^{Sassaki et al., 2018}).

Given the current scenario of economic crisis and the advances that have already been made through top-down policies, it seems useful to seek different and complementary strategies for public basic education. Factors at school level related to management, pedagogical practices, autonomy (^{Bloom et al, 2015}) and integration with the community (^{Cunha et al, 2017}) have a significant influence on student performance. Therefore, public management and, in particular, the regulation of education, must adopt strategies that are complementary to the centralizing regulatory instruments that have characterized advances in education in recent years.

The need for a broader range of regulatory strategies also arises from the multiplicity of situations that make up the problem of education in Brazil, since there are different stages of development not only between states, but also between cities and, even more so, between schools that make up the education network of the same city. It is reasonable to think that a certain degree of infrastructure (including access to meals, school transportation and structural conditions such as libraries, laboratories and other technological structures), a training plan for teachers, the supply of teaching materials and the existence of a minimally adequate curriculum are necessary conditions for schools to achieve better results. After a certain point, however, these instruments have less impact on learning outcomes, which is why the opportunity cost of different possible regulatory strategies should be taken into consideration in order to optimize the process of improving learning indicators.

To demonstrate the usefulness of complementary strategies to the regulatory instruments already being used by the Brazilian Federal Government, this article will analyze the following hypotheses: (i) Schools in the same municipal school network (in other words, schools from the same city) show relevant IDEB differences; (ii) regulatory instruments based on a bottom-up strategy could be used in a complementary way to the policies already implemented to improve the learning outcomes of students.

The city of Rio de Janeiro will be used as a test case for the above hypotheses. This article will use georeferenced data from the city³³ on the performance of municipal schools in IDEB³⁴, with specific target aiming at the early years of elementary school (Elementary I). These data make it possible to check whether there are neighboring - and therefore comparable - schools with very discrepant results. The information collected by the INEP³⁵ and the Rio de Janeiro City Hall’s Department of Education also allows for testing whether any differences between school pairs can be reduced to socio-economic factors, as measured by the Indicator of the Socioeconomic Level of Basic Education - INSE³⁶ and the limits of favelas.

The physical proximity of discrepant school pairs, if detected, is indicative that at least some of the challenges linked to improving the quality of Basic Education need to be addressed with regulatory instruments complementary to the federal measures mentioned, in particular using the bottom-up approach. After all, if it is proven that some schools achieve good results (or relevant developments in the indicators) despite the difficulties they face, their practices can serve as an example to inform recovery measures for other school units that perform worse. Learning lessons from local experiences is what characterizes bottom-up regulatory strategies, and the best way to implement them remains to be seen (whether through help from the Federal Government or through a change in the regulatory governance of cities in order to incorporate programs and/or bottom-up regulatory measures into their portfolio).

The choice of the initial years of elementary school is justified by its probable impact on the later stages (final years of elementary school and high school). The hypothesis is that the deficiencies in teaching in the early years of elementary school impair what needs to be learned by students in the final years and generate serious consequences in the development of the student during high school³⁷. The statistics are consistent with this hypothesis. When looking at math learning alone, for example, only 42.9% of students leave the initial years of Elementary School (Ensino Fundamnetal I) with an adequate level of learning. This number drops to 18.2% at the end of Ensino Fundamental II and 7.3% at the end of High School³⁸ (^{Todos pela educação, 2018}). The same story is repeated for the performance of students in Portuguese, with similar data.

Presumably, students who left the early years of primary school with insufficient ability in the core subjects of the curriculum - Portuguese language and mathematics - have greater difficulty in learning, which can generate not only a cascading effect on student performance over time, with possible impacts on school dropout, due to its effect on the age/student year ratio. If this reasoning has any merit, the most efficient way of allocating resources is sequential, confronting first the deficits that present themselves in the initial years of elementary school and, later, in their final years. The choice of the City of Rio de Janeiro as the target of the investigation was more pragmatic: the city stands out as a reference center in open data (^{Schreiner, 2016}, p. 26).

Differences in Learning Outcomes in the Same School Network: possibility of a bottom-up approach

There are strong indications that bottom-up regulatory strategies can represent significant advances in the quality of learning: the “Excellence with equity” research project (^{Faria; Guimarães, 2015}, Lemann Foundation; Itaú BBA, 2012), which aimed to identify and investigate some reference schools in different regions of Brazil with quantitative and qualitative methods pointed to the importance of qualified principals and good management practices at the school level, drawing attention to peculiarities of teaching that are difficult to regulate through centralizing instruments. Quantitatively, these efforts make a difference: ^{Bloom et al (2015}), in a study that investigated eight countries, including Brazil, showed the high degree of influence exerted by school management practices - in particular, their degree of administrative autonomy - over the academic performance of their students.

Taking into account what has been said above regarding the impact of specific management measures, the need for complementary regulatory instruments makes sense if: (I) there are municipal schools with discrepant IDEB scores in the early years of elementary school and (II) these schools are physically and socioeconomically close. Using data from the Department of Education of the Rio de Janeiro City Hall (^{PCRJ, 2019}) and microdata from IDEB (2019), it was possible to explore these variables, creating a heat map where each point represents a school in the municipal education network and the color temperature represents the school’s IDEB score. The warmer the color (the closer to black), the smaller the school’s IDEB score.

Source: authors,based on data from IDEB (2017) and Rio de Janeiro City Hall (PCRJ, 2019)

Figure 1 IDEB in Basic Education - Elementary School - Initial Years - of Rio de Janeiro’s Municipal Schools³⁹  

Figure 1 shows that both assumptions seem plausible. In a manner consistent with the specialized literature regarding the factors that influence school performance (^{Alves; Soares, 2007a}, Alves; Soares, 2013b; ^{Andrade; Laros, 2007}; ^{Gramani, 2017}), some areas of the city have a higher concentration of schools with low performance, corroborating already known hypotheses about the weight of socioeconomic issues in the learning outcomes of children and young people. However, in practically all districts of the city of Rio de Janeiro, it is possible to observe much lighter (orange) points (higher IDEB scores) alongside darker (black) points (lower IDEB scores), indicating that in fact there are neighboring schools with substantially different performance.

On the existence of Reference Schools and Underperforming Schools in the School Network of the City of Rio de Janeiro - RJ

Which schools can be considered a reference in this context? If the goal is to find the variables at the school level that can be manipulated to make schools with poor performance reach an acceptable level of quality, we should ideally identify the schools that have gone through this history. Thus, instead of focusing solely and exclusively on school performance in the last IDEB assessment - 2017 - we decided to select the schools that showed an expressive evolution in IDEB scores between the measurements of 2005 and 2017. Two different categories were created for the reference schools: 1) Schools that went from very bad performance to good performance: in this first classification are the schools that had IDEB scores below 4 in the 2005 assessment and reached IDEB scores above 6 in the 2017 assessment (24 schools); 2) Schools that went from bad performance to good performance: similarly to the first category, schools that had IDEB scores below 5 in the 2005 assessment and IDEB scores above 6 in 2017 (124 schools) were included in this group.

The other side of the problem also demands a narrow definition (i.e. what criterion would serve to identify underperforming schools). With this in mind, underperforming schools were defined as those that simultaneously fulfilled two conditions in the city of Rio de Janeiro: 1) Schools which occupied the last quartile of IDEB scores in the 2017 evaluation (159 schools); and 2) Schools which occupied the last quartile of the average for each school of the last 4 IDEB measurements (2011, 2013, 2015 and 2017) - 135 schools.

Either criterion, applied in isolation, would be potentially misleading. The first because, while it may capture which schools are performing worse at the moment, it may also include traditionally excellent schools that, because of some merely conjectural reason, performed poorly in the 2017 IDEB. The second criterion could leave out schools that showed some improvement in the last IDEB, even though they had difficulties in the first evaluations of our timeframe.

The joint application of the criteria identified 103 schools that need to be assisted through state policies. After the exclusion of two schools with incomplete data⁴⁰, a universe of 101 schools remained in this category (underperforming schools, recipients of aid).

Figure 2 represents, for each underperforming school, the closest reference school, allowing a visualization of their geographical positions:

Source: the authors, based on data from IDEB (2005-2017) and Rio de Janeiro City Hall (PCRJ, 2019)

Figure 2 Underperforming schools and reference schools 

Figure 2 shows the existence of reference schools alongside underperforming schools⁴¹. In the most emblematic cases for performance comparison purposes, the nearest reference school is only a few hundred meters from the underperforming school (see Table 1).

In a few cases, the distance between the underperforming school and the reference school is large (see Table 2), but these distances may still not represent relevant comparative challenges for the two schools.

Table 1: 10 shorter distances between schools 

Underperforming school	Reference School	Distance (meters)
CIEP THOMAS JEFFERSON42	LIMA BARRETO MUNICIPAL SCHOOL	98.083
ALBINO SOUZA CRUZ MUNICIPAL SCHOOL	EMA NEGRAO DE LIMA MUNICIPAL SCHOOL	103.77
JENNY GOMES MUNICIPAL SCHOOL	PEREIRA PASSOS MUNICIPAL SCHOOL	246.9
MUNICIPAL SCHOOL RUBENS BERARDO	MUNICIPAL SCHOOL PROFESSOR AFONSO VARZEA	253.62
CIEP DOCTOR NELSON HUNGARY	MUNICIPAL SCHOOL PROFESSOR JOAO GUALBE...	297.48
MUNICIPAL SCHOOL ARAUJO PORTO ALEGRE	MARIO FACCINI MUNICIPAL SCHOOL	369.22
MUNICIPAL SCHOOL PRESIDENT E JOAO GOULART	PANAMA MUNICIPAL SCHOOL	425.55
MUNICIPAL SCHOOL SOARES PEREIRA	MUNICIPAL SCHOOL BARAO DE ITACURUSSA	428.06
CIEP COLONEL SARMENTO	EURICO VILLELA MUNICIPAL SCHOOL	448.42
MEM DE SA MUNICIPAL SCHOOL	PEREIRA PASSOS MUNICIPAL SCHOOL	469.14

Source: Own elaboration, based on data from IDEB (2005-2017) and Rio de Janeiro City Hall (PCRJ, 2019).

Table 2: 10 longer distances between schools 

Underperforming school	Reference School	Distance (meters)
CIEP DEPUTY ULYSSES GUIMARAES	MUNICIPAL SCHOOL TEACHER MARIA SANT...	5685.05
MUNICIPAL SCHOOL TEACHER CASTILHO	MUNICIPAL SCHOOL TEACHING LEOCADIA T...	5333.84
CIEP VILA KENNEDY	SAMUEL WAINER MUNICIPAL SCHOOL	3075.76
JOANA ANGELICA MUNICIPAL SCHOOL	SAMUEL WAINER MUNICIPAL SCHOOL	3058.39
MUNICIPAL SCHOOL DRIVER PASCHOAL ANDRE	MANUEL DE ABREU MUNICIPAL SCHOOL	3014.43
CIEP ANTONIO CANDEIA FILHO	ISAIAS ALVES MUNICIPAL SCHOOL	2993.62
JULIO CESARIO DE MELO MUNICIPAL SCHOOL	MUNICIPAL SCHOOL TEACHER MARIA SANT...	2944.2
ALZIRO ZARUR MUNICIPAL SCHOOL	MANUEL DE ABREU MUNICIPAL SCHOOL	2943.04
MUNICIPAL SCHOOL PADRE JOSE MAURICIO	RAYMUNDO CORREA MUNICIPAL SCHOOL	2939.46
GENERAL MUNICIPAL SCHOOL OSORIO	ISAIAS ALVES MUNICIPAL SCHOOL	2789.42

Source: The authors, based on data from IDEB (2005-2017) and Rio de Janeiro City Hall (PCRJ, 2019).

Thus, it is possible to observe that the data supports the hypotheses raised at the beginning of the paper. It was possible to notice that schools within the municipal network of Rio de Janeiro present disparate academic results and that these differences are reproduced even in geographically close pairs. Finally, we found that the distance between reference schools and underperforming schools is often quite small and that, even in the most dramatic cases, we found reference schools at a reasonable distance from each underperforming school.

Difference in Performance Correlated to Socioeconomic Factors

The motivation for selecting the closest school pairs was the assumption that geographically nearby schools are similar in relation to the top-down variables (i.e., they are eligible for the same funding programs, are subject to the exact same curriculum, have the same career plan and training-level for teachers, among others). In addition, district-level HDI and GDP, for example, are likely to remain constant for neighboring schools since, as can be seen from Table 1, these schools are usually in the same neighborhood or in immediately neighboring neighborhoods. On the other hand, one of the most important variables for school performance, the socioeconomic level of students, may vary among neighboring schools, which could explain the difference in results among the pairs of schools identified, despite their physical proximity (as a function of the difference among students). Therefore, an important question arises here: can the differences between reference and underperforming schools be explained by this factor?

We used data from INSE (^{INEP, 2015b})⁴³ to test this hypothesis. Although there is a correlation between INSE and IDEB for the 149 schools evaluated at this stage of the study⁴⁴, the result does not replicate for school pairs. When looking only at the distance between the INSE of the reference school and the underperforming school and the difference in IDEB performance between the school pair, there is no correlation⁴⁵, as can be seen in Figure 3.

This result can be interpreted as follows: although certain regions have a higher concentration of school-references and there is an influence of INSE on absolute IDEB scores for the neighborhood, the physical proximity of school pairs implies a proximity in the student body, so that the variation among pairs cannot be explained only with reference to the socioeconomic level of students.

Source: IDEB (2017); INSE (2015)

Figure 3 Correlation between IDEB (2017) and INSE (2015) differences for each closer school pair⁴⁶  

From the data presented in Figure 3, it can be seen that: (i) There is a high number of underperforming schools in the city of Rio de Janeiro; (ii) For most of these schools, there is a reference school at a reasonable distance that can serve as an example (for the design of management practices that worked in a comparable context).

As we will argue in the next few pages, this information is extremely relevant for the regulation and improvement of municipal education. Moreover, socioeconomic variations among students in each school pair have not been able to explain the difference in performance of these students, which is suggestive of the effectiveness of local management practices implemented in the reference schools⁴⁷.

Regulatory Solutions and Principal/Agent Problem: Efficiency of Local Regulatory Instruments

Since 2007, the Ministry of Education (Ministério da Educação e Cultura - MEC) has created a series of instruments with the objective of encouraging and assisting local educational policies, both at State and Municipal levels⁴⁸. In particular, a series of actions were developed under the Education Development Plan (Plano de Desenvolvimento da Educação - PDE), with the objective of fulfilling the All for Education Commitment goal plan, incorporated into Brazilian law by the Decree 6.094/2007, to which all Brazilian municipalities adhered (or at least the majority, see ^{Souza; Alcântara; Vasconcelos, 2014}, p. 225). The main decentralized form of incentive for the development of education under the PDE is the possibility for cities of making agreements with Federal and State governments involving federal funds and technical assistance.

In order to adhere to the PDE and become eligible for the financial and technical transfers, each city had to analyze, in light of the criteria established by the MEC, the performance of its own education and, based on this diagnosis, prepare an Articulated Action Plan (Plano de Ações Articuladas - PAR), which serves as a guideline for the celebration of eventual agreements (^{Souza; Alcântara; Vasconcelos, 2014}; ^{Oliveira, 2014}, p. 413). In practice, the informatics system in which cities must register their diagnosis automatically suggests the actions that must be implemented by the Department of Education based on the answers provided to the evaluation items⁴⁹. Although formal adherence to the PAR has been high, particularly in the State of Rio de Janeiro, the diagnostic step is subverted by perverse incentives. Souza, Alcântara and Vasconcelos (2014), for example, published the research paper “Local Diagnostics of the articulated action plan: an analysis of its reliability” in which they explore the contradiction between the IDEB score of the State of Rio de Janeiro in 2011 - one of the 5 worst in Brazil - and the positive evaluation that the cities made in the PAR diagnosis.

According to the authors cited, the reasons for these discrepancies may be linked to the asymmetry of information combined with discrepant interests:

[...] because, by attributing higher scores [the cities] find themselves free from the description of the actions that the PAR system thinks should be adopted, [in addition to avoiding] [...] greater regulation by federal government policies (since, in the case of unsatisfactory evaluations, the municipality has to face up to the consequences stipulated in the plan) (^{Souza; Alcântara; Vasconcelos, 2014}, p. 228).

Thus, on a number of significant occasions, the PAR does not seem to meet its decentralizing objectives⁵⁰. This becomes clear when one notices that the automatic coupling of actions to diagnosis causes the perverse incentives described above to have practical consequences.

From the perspective of the economic analysis of law, the problems faced in implementing PAR appear to be principal-agent problems (^{Melo, 1996}), in which the interests of the public (principal) are not achieved by those responsible for city level education (Municipal Departments of Education and, ultimately, school principals). As this is an information problem, there is a need to build local regulatory instruments to increase transparency and accountability⁵¹ on the measures needed to repair underperformance in the schools identified⁵². And because of the principal-agent problem, compounded by possible additional costs associated with political arrangements (mayors and governors may be from distinct political parties, and may still be politically opposed to the holders of office in the federal government), federal solutions are likely to be unable to effectively eliminate the performance gap.

It seems, therefore, that the creation of local regulatory governance responsible for reparatory actions is the most efficient way to complement the top-down approach, which has effectively improved results over the past 15 years. To reverse the underperformance of the identified schools, it is necessary to look at a series of factors specific to each of the units, including the conduct of school principals and the way they relate to teachers (^{Bloom et al., 2015}, Lemann Foundation; Itaú BBA, 2012); the conduct of members of state regulatory bodies and their contact with principals and, finally, the relationship between the parents of students and the school (^{Cunha et al., 2017}; ^{Coelho, 2008}).

Decentralizing policies have been adopted in several countries, with positive results (see ^{Bloom et al, 2015}). Examples of this international trend for valuing bottom-up strategies can be found in the United Kingdom, with the academies program - academies are basic education institutions, completely independent of public authorities (Eyles; Machin, 2018) - and in Portugal, where administrative decentralization of the public network is indicated as one of the reasons for the country’s improved performance on PISA (Liebowitz et al., 2015). Even the US, which had a controversial decentralizing experience with the No Child Left Behind program, still signals towards the valorization of local policies, for example, by returning the regulation of incentives linked to performance evaluations to Sates as a solution to their educational problems (^{Sharp, 2016}).

The largest success story in Brazil, widely used as an example in the literature, is that of the city of Sobral, in Ceará, which went from an IDEB score of 4.0 in the initial years of elementary school in 2005 to a score of 9.1 in 2017 (^{Gramani, 2017}; ^{Menezes Filho, 2015}; Inep, 2005a). Since 2001 (^{Inep, 2005a}), the aforementioned city has adopted management strategies that seek to eradicate illiteracy and school evasion by valuing teachers, establishing performance-based bonuses and awards, and measuring the results every six months with the aid of an external commission. The State of Ceará as a whole has stood out in Basic Education as a result of strategies similar to those employed in Sobral, which emphasize the local character of education, although there is still considerable room for improvement (Gramani, 2017, ^{Abrucio; Seggatto; Pereira, 2018}).

In Rio de Janeiro, however, there is no regulatory instrument that organizes reparatory actions for underperforming schools, bringing transparency to necessary actions, learning outcome targets and even incentives or sanctions in the event of non-compliance. There are important provisions regarding Basic Education in the Municipal Organic Law (the relevant city-level statute), which, after repeating many of the provisions already present in the federal Constitution, delegates the specific regulation of the subject to the Municipal Education Plan (lano Municipal de Educação - PME) which, however, does not provide a regulatory instrument that is capable of guiding a bottom-up strategy. In particular, large cities, such as the City of Rio de Janeiro, with a very large number of schools⁵³, need regulatory governance that allows greater transparency for each of their schools, as well as for the planning of individualized actions focused on underperforming schools, which will only be possible with empirical evaluations such as the one presented in the first section of this paper.

The latest version of the PME in Rio de Janeiro was approved by Law 6.362/2018. Structured by means of sets of goals and strategies, the PME establishes concrete objectives regarding the expansion of access and quality of Basic Education, including target scores in IDEB and PISA. However, there is no provision for specific reparatory actions⁵⁴, which need to relate directly to the existing discrepancies between schools in the municipal education network, although there are local strategies to address the learning problem⁵⁵. There is also no sign of publicly funded research to identify practices that make certain schools, principals and teachers stand out from the rest. Adopting PME-based programs as they stand today may not be enough to address the difficulties presented by underperforming schools, as they do not allow for the individualized actions typical of a bottom-up approach to implementing public policies, nor do they encourage strategies that prioritize more effective results in the quality of learning, such as a reparatory focus in the early years of elementary school.

Conclusion

The analysis of the educational data of the City of Rio de Janeiro showed that there are neighboring schools with disparate performances in standardized tests. There are pairs of schools with underperforming and reference schools less than one kilometer apart. In addition, socioeconomic differences among students in these schools cannot explain the difference in their performance. However, currently the set of regulatory instruments used to address reparatory actions is insufficient, either because of the principal-agent problems in accessing federal regulatory instruments or because of the absence of regulatory mechanisms at the level of the City of Rio de Janeiro in this direction.

In general, there is an opportunity for the implementation of decentralizing regulatory policies, which complement top-down strategies with bottom-up ones. Finding a regulatory solution for education in the city of Rio de Janeiro therefore requires detailed consideration of local characteristics. The present work, lacking a qualitative empirical step, limited as it is to the quantitative identification of school pairs that serve as confirmation for the need of a bottom-up strategy, is not sufficient to find a solution. It would be inherently contradictory to preach the importance of local variables and go on to outline a solution for Rio de Janeiro solely based on the strategies that worked, for example, in the UK. However, it is already possible to advance some conclusions.

The first is the one resulting from the empirical analysis carried out in this paper: it is necessary to develop bottom-up regulatory instruments for the city of Rio de Janeiro. To the extent that local variables not associated with socioeconomic factors matter significantly for school performance, the need to develop regulatory instruments that exert local influence and are capable of inducing the behavior of agents seems obvious. The relative lack of regulation at the municipal level - observed in the case of the city of Rio de Janeiro - is not justifiable in view of the convergence of empirical and conceptual evidence around the importance of school management. The improvement in data requires an improvement in public management (Ragazzo, 2011).

Moreover, given the good outcomes of top-down policies, it seems reasonable that the focus of local regulatory instruments should be on reversing the trend on the worst performing schools. The idea that public education should be offered with equity finds a lot of support in literature (^{Faria; Guimarães, 2015}, Lemann Foundation; Itaú BBA, 2012, ^{Comunitas; Juntos pelo Desenvolvimento Sustentável; FGV Ebape, 2017}, ^{Coelho, 2008}). Social justice require that the focus of state efforts should be in underperforming schools, especially for the first years of elementary school, since that learning at this stage will have impacts on the later stages of students’ education.

The existence of reference schools throughout the city of Rio de Janeiro shows that it is possible to overcome the financial, administrative and structural limitations that afflict public education and to achieve good results, which suggests that bottom-up strategies can represent a complementary approach in order to understand how reference schools’ management practices within their geographical area can shorten the path to identifying the solution to repair the results of underperforming schools.

The results of the analyses carried out in the research also make clear that these solutions have to be designed by local agents, due to the principal/agent problem that arises from situations in which the aid depends on access to the Federal Government, which, by the way, already occurs through the PAR and which, until now, has not proved sufficient to revert the situation identified in the City of Rio de Janeiro, which has over 100 underperforming schools.

The task carried out in this article justifies the struggle to maintain and improve instruments such as IDEB and the School Census, because only with reliable and consistent data sets will it be possible to adequately evaluate education. This data enables the development of a local regulatory governance capable of creating additional regulatory instruments for the monitoring of school performances (with purposes and periodicity distinct from federal assessment instruments), repair and aid, providing greater transparency and accountability for government actions addressed at the educational deficits of Brazilian cities.

Finally, it is important to point out that Rio de Janeiro (as well as Brazil) has already demonstrated through concrete examples that it is capable of high quality education, since the municipal school with the best performance in elementary school has an IDEB score of 8.0⁵⁶. Therefore, it is possible to achieve excellence. The difficulty is to extend this knowledge to schools with low performance, generating scale by learning the management techniques that have a positive impact on IDEB scores. New regulatory instruments, such as the one suggested in this paper based on the identification of pairs of reference and underperforming schools for the dissemination of best practices, can be applied to cities with profiles similar to that of Rio de Janeiro, thus maximizing results, minimizing costs and improving indicators, but with different impacts on municipal schools, demanding individualized reparatory actions for a relevant number of units on a given educational network.