• Nenhum resultado encontrado

School feeding in Brazil: a Cost Assessment of the National Program

N/A
N/A
Protected

Academic year: 2021

Share "School feeding in Brazil: a Cost Assessment of the National Program"

Copied!
82
0
0

Texto

(1)

School Feeding in Brazil:

A Cost Assessment of the National Program

(2)

1

Index

Abbreviations and Acronyms ... 3

List of Tables ... 4

List of Figures ... 6

Abstract ... 7

Executive Summary ... 8

1. Introduction and motivation ... 10

2. Brazilian School Feeding Funding Process... 12

3. Databases and other sources of information ... 16

3.1. Secondary Data ... 16

3.2. Primary Data ... 18

4. Expected Outcome ... 19

5. Methodology ... 20

5.1. Meals Served ... 20

5.2. Total Food Costs ... 21

5.2.1. Federal Transfers ... 21

5.2.2. Local Transfers for Food (LTF) ... 22

5.3. Infra-Structure & Other Costs ... 24

5.3.1. Defining Representative School Costs Profiles ... 24

5.3.2. Defining an Input-Usage Function for Each Representative Profile ... 26

5.3.3. Pricing Inputs of the Representative Input-Usage Function ... 27

5.3.4. Estimating Infra-Structure & Other Costs for each Brazilian School ... 27

5.4. Estimating Total Meal Cost per School ... 27

6. Data Collection Process ... 29

6.1. Online Survey ... 29

6.1.1. Answers from Secretaries ... 30

6.1.2. Answers from Municipalities ... 31

6.2. Interviews: Telephone and In-person ... 32

7. Empirical Results ... 34

7.1. Descriptive Statistics ... 34

7.1.1. Main Costs ... 34

(3)

2

7.2. Cost Assessment and Predictions ... 43

8. Final considerations ... 2

9. References ... 3

10. Annexes ... 4

10.1. Descriptive Results in Reais ... 4

Counterpart Budget Distribution for Secretaries (in R$ of 2013) (n=2) ... 4

Average Monthly Secretariat Cost per Meal, by Socioeconomic Condition and Cost Category ... 4

Cost per Meal in each Group Classification by Region, per Socioeconomic Condition ... 5

Average Meal Cost per Group of Socioeconomic Condition and Region, for Small Sized Schools ... 5

Average Meal Cost per Group of Socioeconomic Condition and Region, for Average Sized Schools ... 6

Average Meal Cost per Group of Socioeconomic Condition and Region, for Big Sized Schools ... 6

Detailed Cost Composition for Secretariats by Socioeconomic Condition – Cost Categories ... 1

Detailed Cost Composition for Secretariats by Socioeconomic Condition – Cost per Meal... 2

Detailed Cost Composition for Schools by Socioeconomic Condition ... 3

10.2. Predicted Costs under both Econometric Strategies and in Reais ... 6

Predicted Cost per Meal for Brazil, under specifications 1 and 2 ... 6

Summary statistics for the Predicted Cost per Meal by specification, per state and region ... 6

Summary Statistics for the Predicted Cost (in US$, 2103) by Specification Strategy, per Socioeconomic Condition ... 8

10.3. Translated Questionnaires ... 9

School Level School Feeding Program Cost Analysis Questionnaire ... 9

Secretariat Level School Feeding Program Cost Analysis Questionnaire ... 16

(4)

3

Abbreviations and Acronyms

CAE Conselho de Alimentação Escolar School Feeding Council

CAPA

Centro de Apoio ao Pequeno

Agricultor Support Center for Small Farmers

CECANE

Centro de Colaboração em Alimentação e Nutrição Escolar

Center for Collaboration in School Feeding and Nutrition

C-MICRO /FGV

Centro de Estudos em Microeconomia Aplicada da Fundação Getulio Vargas

Center of Applied Microeconomics Studies at Fundação Getulio Vargas

CONAB

Companhia Nacional de

Abastecimento National Supply Company

CONSEA

Conselho Nacional de Segurança

Alimentar National Council of Food Security

COPASA

Companhia de Saneamento de

Minas Gerais Minas Gerais Sanitation Company

EMATER

Empresa de Assistência Técnica e Extensão Rural

National Company for Technical Assistance and Rural Extension

EPAGRI

Empresa de Pesquisa

Agropecuária e Extensão Rural de Santa Catarina

Santa Catarina State Company for Agricultural Research and Rural Extension

FNDE

Fundo Nacional de

Desenvolvimento da Educação

National Fund for Educational Development

HDI - Human Development Index

IBAMA

Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis

Brazilian Institute for the Environment and Renewable Natural Resources

IBGE

Instituto Brasileiro de Geografia e Estatística

Brazilian Institute of Geography and Statistics

INEP

Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira

National Institute for Educational Studies and Research Anísio Teixeira

IPEA

Instituto de Pesquisa Econômica

Aplicada Institute for Applied Economic Research

LTF - Local Transfers for Food

PAA

Programa de Aquisição de

Alimentos Food Acquisition Program

PNAE

Programa Nacional de

Alimentação Escolar School Feeding Program

SEBRAE

Serviço Brasileiro de Apoio às Micro e Pequenas Empresas

Brazilian Service to Support Micro and Small Enterprises

SENAR

Serviço Nacional de Aprendizagem

Rural National Service of Rural Learning

SESI Serviço Social da Indústria Industry Social Services

UNPD - United Nations Development Program

USD - Dollars of the United States of America

(5)

4

List of Tables

Table 1: Daily per capita federal transfer for a student according to class type (US$ of 2013) ... 13

Table 2: FNDE Education Categories for federal transfers (R$ of 2013) ... 16

Table 3: Sources of Price Consultation ... 18

Table 4: Minimum of meals and Nutritional Needs of those meals that need to be offered to Public School Students per School Day, among FNDE Categories ... 21

Table 5: Distribution of Enrolled Students among School Administrative Spheres ... 22

Table 6: Illustration of the Weighted Proportion of Students on Primary Education of a Hypothetical City ... 23

Table 7: School Feeding Inputs and Their Units of Measure ... 26

Table 8: Econometric Specifications Adopted for Estimating Infra-Structure and Other Costs ... 27

Table 9: Cost Categories on which Counterpart Budget was Spent ... 29

Table 10: Data mining: Cumulative Filters on Survey Responses ... 30

Table 11: Informative and Robust Response Distribution: over Administrative Sphere and Type of Cost Funded by Counterpart Budget (n =151) ... 30

Table 12: Counterpart Budget Distribution for Secretaries (in U$ of 2013) (n=2) ... 31

Table 13: Distribution of Municipal Responses by School Size (n =149) ... 31

Table 14: Distribution of Administrative Responses by School Size and Type of Cost Funded by the Counterpart Budget (n =151) ... 32

Table 15: Average Monthly Secretariat Cost per Meal, by Socioeconomic Condition and Cost Category ... 34

Table 16: Cost per Meal in each Group Classification by Region, per Socioeconomic

Condition ... 36

Table 17: Level of Secretariat Cost per Meal ... 36

Table 18: Average Meal Cost per Group of Socioeconomic Condition and Region, for Small Sized Schools ... 39

Table 19: Average Meal Cost per Group of Socioeconomic Condition and Region, for

Average Sized Schools ... 40

Table 20: Average Meal Cost per Group of Socioeconomic Condition and Region, for Big Sized Schools ... 40

(6)

5 Table 21: Perception of the Secondary Agencies by Secretariats ... 42

Table 22: Perception of the Secondary Agencies by Schools ... 42

Table 23: Regression models for the Cost per Meal in Brazilian Municipal Elementary

Schools, under OLS specification 1 ... 44

Table 24: Regression models for the cost per meal in Brazilian Municipal Elementary

Schools, under OLS specification 2 ... 45

Table 25: Predicted Cost per Meal (in US $ 2013) for Brazil, under specifications 1 and 2 ... 45

Table 26: Summary statistics for the Predicted Cost per Meal (in US $ 2013) by Specification, per State and Region ... 47

Table 27: Summary Statistics for the Predicted Cost (in US$, 2103) by Specification Strategy, per Socioeconomic Condition ... 50

(7)

6

List of Figures

Figure 1: Dimensions of a School Class and Eligibility for Public Funding for Food (categories marked in red are not eligible to PNAE) ... 12

Figure 2: Resources used for School Feeding ... 14

Figure 3: Inputs for PNAE and the Origin of the Funds Used ... 14

Figure 4: Three Determinant Dimensions in Brazilian School Feeding Cost Structure and its features ... 25

Figure 5: Distribution of Secretariat Cost per School by cost category, per Socioeconomic Condition ... 35

Figure 6: Number of Secretariats by Cost Level per Region ... 37

Figure 7: Average Monthly Costs of Schools, by school size ... 38

Figure 8: Predicted Costs per Meal (in US$ 2013) for States and Brazil, under OLS

specification 1 ... 48

Figure 9: Predicted Costs per Meal (in US$ 2013) for States and Brazil, under OLS

specification 2 ... 49

Figure 10: Predicted Cost per Meal (in US $2013), by Socioeconomic Condition under OLS specification 1 ... 51

Figure 11: Predicted Cost per Meal (in US $2013), by Socioeconomic Condition under OLS specification 2 ... 52

(8)

7

Abstract

The Brazilian National School Feeding Program (PNAE) plays an important role in Brazilian National Education and given its broad structure and context, may serve as a guide for other developing countries. However, due to its large scale and its complexities, PNAE’s cost structure has not been thoroughly investigated. This study proposes a methodology to estimate the total cost for delivering a meal to a primary school student in Brazil1. The results indicate that the southern regions (south and southeast) present higher cost per meal: US $ 1.55 and US $1.30 under econometric specification 1 and 2 respectively. Furthermore, the northern regions (northeast and north) present the lowest cost per meal: US $ 0.89 and US $ 0.90 under econometric specification 1 and 2 respectively. Furthermore, schools in areas of medium socioeconomic condition present higher cost per meal than those in areas of low and high socioeconomic condition. The average predicted cost per meal under OLS specification 1 by socioeconomic cluster (low, medium and high) is, respectively, US $ 1.03, US$ 1.31, and US $ 1.19. The average predicted cost per meal under OLS specification 2 by socioeconomic cluster is, respectively, US $ 1.13, US$ 1.30, and US $ 1.03.

Key words: Cost Assessment; National School Feeding Program; Primary Education;

Brazil

1This study is supported by the WFP through Bill & Melinda Gates Foundation, whose main mission is to augment populations’

welfare through actions of enhancing healthcare, fighting poverty and promoting educational opportunities and access to information technology. In that sense, it is interesting to comprehend the school feeding cost structure in Brazil, since this country has a universal school feeding program and then it may serve as a role model to other underdevelopment countries.

(9)

8

Executive Summary

Brazil´s Programa Nacional de Alimentação Escolar - PNAE is the second largest National School Feeding Program worldwide and the largest among the universal public programs. By 2014, the program was responsible for feeding over 43 million students in the Country, and relied on a budget of USD 1.35 billion.

The present study aims to provide a methodology for the cost assessment of PNAE, seeking to carefully estimate the total cost per meal delivered to a primary education student in Brazil, and its 27 states2. With this, the study seeks to illustrate part of the existing cost heterogeneity in Brazil among several dimensions regarding socioeconomic development conditions.

This study may be useful for the revision of the Program as a whole, in that it may provide empirical feedback to the Brazilian government, leading to a better understanding of the program and its transmission mechanisms. Furthermore, it may also serve as a guideline for other developing countries aiming at implementing or expanding similar programs. The mission of the World Food Program is to end global hunger.

The lack of detailed data is the main challenge to precisely specify the cost structure of PNAE. This study will address this challenge by gathering data from different surveys and databases, collecting primary data and implementing statistical methods for predictions and inference using all the available information.

The school feeding costs may be separated into two major groups - Food and Infra-Structure & Other Inputs3. For the first group (Food Expenses) the aggregate values by municipality are available and amount spent on primary education (under a few assumptions listed below) is easily obtained. However, the second group (Non-food Expenses), there is very limited knowledge of components and must be dealt with by this study.

The methodology delivers one final number, which is the average cost for delivering a meal on Primary School in Brazil in the year of 2013, as well as average costs for the 27 Brazilian states. The cost will be estimated by diving total food costs and infrastructure and other costs by the total number of meals. This cost will be analyzed in three 3 separate steps, namely: (i) meals served; (ii) total food costs; and (iii) infra-structure & other costs.

Using two specifications under an Ordinary Least Square econometric strategy, the total cost per meal was predicted. Brazil´s heterogeneous socioeconomic conditions lead to a need for assessing state-by-state cost per meal.

The predicted costs under the OLS strategy using both specifications (1 and 2) are US$ 1.03 and US$0.95, respectively. The results found per state are heterogeneous: under specification 1 the southeast presents the highest cost per meal (US $1.55), while under specification 2, the south presents the highest cost per meal (US $ 1.30). Under specification

2There are approximately 54% of Brazilian students enrolled in the primary education (School Census, 2012).

3Although cost structure is divided into two major groups for the sake of organizing the methodology, the cost structure

(10)

9 1, the northeast presents the lowest cost per meal (US $ 0.89) while the north assumes this place under specification 2 (US $ 0.90). Nonetheless, under both specifications, the southern regions present higher cost per meal, while the northern regions present lower cost per meal.

Furthermore, when decomposing the results by socioeconomic condition the medium socioeconomic cluster presents highest predicted cost per meal (US $ 1.31 and US$ 1.30, respectively). The average predicted cost per meal under OLS specification 1 by socioeconomic cluster (low, medium and high) is, respectively, US $ 1.03, US$ 1.31, and US $ 1.19. The average predicted cost per meal under OLS specification 2 by socioeconomic cluster is, respectively, US $ 1.13, US$ 1.30, and US $ 1.03. Under both OLS specifications adopted (1 and 2).

(11)

10

1. Introduction and motivation

Brazil´s Programa Nacional de Alimentação Escolar - PNAE is the second largest National School Feeding Program worldwide and the largest among the universal public programs. The program was created in the 50's and reformulated in 1994 under the mandate of President Cardoso. At this time, school food purchases were decentralized. Since its establishment, PNAE has contributed to public school students’ development, thus contributing to their learning process and cognitive development.

By 2014, the program was feeding around 43 million students and had a very complex budget structure of approximately USD 1.35 billion per year. The program is funded mainly by the federal government and complemented with funds from municipalities and state governments, as a result of the decentralization of the program.

The present study aims to provide a methodology for the cost assessment of PNAE, seeking to carefully estimate the total cost per meal delivered to a primary education student in Brazil4. This study may be useful for the revision of the Program as a whole, in that it may provide empirical feedback to the Brazilian government, leading to a better understanding of the program and its transmission mechanisms. Furthermore, it may also serve as a guideline for other developing countries aiming at implementing or expanding similar programs.

The mission of the World Food Program is to end global hunger. As a direct result, one of WFP actions is to improve the quality of life of vulnerable populations, mainly by improving eating habits and food sustainability. Various studies5 have shown that quality of food, rather than simply caloric intake, have a statistically significant and positive impact on children cognitive skills. As a result, school feeding programs are crucial to enhance and guarantee quality of life.

School feeding and/or take-home ration programs are coordinated by WFP in over 71 countries, and increased school attendance by creating incentives for parents to enroll their children and delivering some basic conditions for learning. As some of the countries where WFP is acting on have socio-demographic similarities to Brazil, the study of PNAE’s cost structure may serve as a useful guideline for WFP managerial plans on national feeding programs worldwide.

The present document seeks to present and execute a methodology for cost assessment of PNAE. Based on several different datasets, this methodology proposes a statistical model to estimate school feeding costs in regard to several school characteristics. This study is organized in seven sections: section 2 briefly presents the funding structure for school feeding in Brazil; section 3 describes the databases used throughout this investigation; section 4 displays the main expected outcome to be reached with this study; section 5 details the methodological approach for cost estimation, highlighting the methods and assumptions used in order to calculate the expected result; section 6 describes the data collection process;

4There are approximately 54% of Brazilian students enrolled in the primary education (School Census, 2012).

5Whaley, Shannon E., et al. "The impact of dietary intervention on the cognitive development of Kenyan school children." The

(12)

11 section 7 presents the estimation results, including descriptive statistics and outcomes of the proposed statistical model; and finally, section 8 contains the final considerations of the study.

(13)

12

2. Brazilian School Feeding Funding Process

Brazil has a dual educational system, with private and public providers. The private system accounts for nearly 8.5 million students, while the public system accounts for approximately 46.3 million students (majority of enrolled students). In private schools, students usually purchase their own food in the cafeterias. However, in public schools6, the government has developed the National School Feeding Program (Programa Nacional de Alimentação Escolar – PNAE), a federal program with the purpose of providing free of charge food for students.

Due to the complexities of the Brazilian National Education System, Brazilian government uses several dimensions to categorize a school class, which are useful for the management of PNAE. A school class can be classified to three main categories of dimensions: (i) education type; (ii) school´s administrative sphere; and (iii) educational level. Figure 1 below illustrates the different possible categories of a school class and how they relate to the eligibility of PNAE, i.e. if the school is eligible to receive public funding for feeding.

Figure 1: Dimensions of a School Class and Eligibility for Public Funding for Food (categories marked in red are not eligible to PNAE)

Source: School Census (2012).

The National Fund for Educational Development (Fundo Nacional do Desenvolvimento da Educação – FNDE), is an autarchy inside the Ministry of Education. It is responsible for the management of PNAE, as determined by Resolution 267, 2013. This recent motion consolidated previous legislation referring to the national school feeding rights and rules. This resolution states that, the provision of food in public schools (and also in non-public

6Eligible schools for public funding are not only public schools, but also private schools in agreement with the government. 7The original name of this Resolution is: Resolução CD/FNDE n° 26, de 2013 Art.38.

(14)

13 schools in agreement with the public sector) is to be funded with both federal and local government resources.

In line with Resolution 26, the amount of federal resources a public school receives consists of a fixed cash transfer per each student, intended exclusively for food purchasing8. The yearly amount of these resources transferred to a given school is obtained by the following mathematical rule:

𝐹𝑅 = 𝑆 × 𝐷 × 𝑉 Where:

FR is the amount of federal resources yearly transferred to a given school; S is the number of students enrolled in the school;

D is the quantity of school days in the year (Resolution 26th fix this number on 200 days);

V is the average daily per capita federal transfer for each enrolled student;

However, it is important to note that per capita value of federal transfers varies according to the type of class or educational program the student is enrolled in, as listed in Table 1 below.

Table 1: Daily per capita federal transfer for a student according to class type (US$ of 2013)9

Student enrolled in

Per capita value

Nursery School 0.43

Kindergarten 0.21

Primary School, Secondary School, Youth/Adult Education 0.13

Indigenous or Quilombola Education 0.26

Full Time Education 0.43

"Mais Educação" Educational Program 0.38

Specialized Education 0.21

Source: Resolution 26th (FNDE, 2013)

Even though federal funding rules are clear and the transfer amount allocated specifically for PNAE can be easily calculated, local government transfers for the National School Feeding Program do not follow specific rules. Therefore, it is not possible to define the total amount transferred by local government a priori. Resolution 26 states only that the local government transfers – or counterpart transfers – are complementary to the federal transfer10.

8Additionally, the Resolution 26th states that 30% of total federal transfer on PNAE must be spent on food from familiar farming.

It’s a policy intended to stimulate social and economic development on local communities.

9To convert the values in Dollar USA (USD), one used the exchange rate ruling on the final period of 2013: 1 DOLLAR USA /

USD = 2.34 REAL BRAZIL / BRL. All following costs are reported in USD. Tables reporting the values in Real Brazil are available at Annex 10.1. Source: Brazilian Central Bank - http://www4.bcb.gov.br/pec/taxas/port/ptaxnpesq.asp?id=txcotacao.

10More specifically, both local and federal governments have joint responsibility in funding the National School Feeding in

(15)

14 Although all the resources listed above through pecuniary transfers finance a significant part of the costs of PNAE, namely local and federal resources, there are further resources used in the public school feeding process. For instance, in order to deliver a meal to a student, a school may use its own capital structure such as kitchenware, stove, the kitchen itself, etc. This is an additional resource used on the National School Feeding process, referred to in this study as School Resources11. It is interesting to note, however, that there is no legal regulation regarding these resources. Gathering all these elements, we are able to illustrate all the necessary resources for school feeding in the chart below (Figure 2).

Figure 2: Resources used for School Feeding

Source: Resolution 26th (2013)

In contrast with the fact that federal resources must be used only for food purchasing, the school resources are not used to fund any part of the schools’ food bill. If needed, the counterpart transfer must complement the federal resources to afford the food expenses, but it is also commonly used for other costs related to the entire process of student feeding, such as food transportation, food preparation, human resources with food production and managerial staff, among others. Thus, for the sake of illustration, dividing the PNAE cost structure roughly into two major parts, namely Food and Infra-Structure & Other Inputs, Figure 3 below illustrated the distribution of each resource type among the two cost categories.

Figure 3: Inputs for PNAE and the Origin of the Funds Used

Source: Resolution 26th (2013)

11It is known that Public School Resources were once acquired with local government budget, so they are in fact funded by

local resources. We are separating it from the Local Resources because they are under different legal aspects (i.e., not mentioned at Resolution 26th) and these are typically physical resources, not monetary transfers.

(16)

15 It is also important to discuss data availability on the flow of the resources. Considering federal resources only, data is available on a yearly frequency on the transfers exclusively for food purchasing per municipality and by all educational levels, so we can precisely obtain the total cost of primary education. For local resources, data is more restricted. The total amount of local resources spent yearly is available for each municipality, and it is possible to separate these among food and non-food expenses. However, it is not possible to disaggregate the non-food items. Therefore, it is not possible to investigate which sub-items and their respective values are contained in the total non-food figure. Lastly, school resource structure is completely unknown, with no available data.

In sum, as previously mentioned, the school feeding costs are separated into two major groups - Food and Infra-Structure & Other Inputs12. The reason for this separation lies on how data will be handled, the base of the methodological rationale. For instance, for the first group (Food Expenses), the aggregate values by municipality are available and the amount spent on primary education (under a few assumptions listed below) is easily obtained. The second group (Non-food Expenses), for which there is very limited knowledge of components, is the black box to be unveiled by this study. The lack of detailed data is the main challenge to precisely specify the cost structure of PNAE. This study will address this challenge by gathering data from different surveys and databases, collecting primary data and implementing statistical methods for predictions and inference using all the available information.

12Although cost structure is divided into two major groups for the sake of organizing the methodology, the cost structure

(17)

16

3. Databases and other sources of information

This section lists and describes the databases and information sources that will be used on the cost assessment.

3.1. Secondary Data

a. School Census (INEP, 2012)

The School Census is a yearly survey covers all public schools in the country, which offer basic education13. The responsible entity for this survey is the National Institute for Educational

Studies and Research Anísio Teixeira (Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira - INEP). INEP is a federal autarchy under the Ministry of Education and is responsible for assessing basic and higher education national.

The School Census collects data that form the basis for the formulation of public policies on national basic education, as well as improving the accountability necessary to the distribution of public resources. Such information (for instance: meals and school transport; books and uniforms; deploying libraries; installation of electricity; Direct Funding for Schools (Dinheiro Direto na Escola), and FUNDEB; number of teachers and teaching positions). Therefore it seeks to monitor the diverse parts that comprise the National Education chain. The year of reference for this database used in the study is 2012.

b. Official Register of Federal Funding of the Ministry of Education (FNDE, 2013)

FNDE registers every federal transfer of PNAE to: (i) municipalities, which manage the municipal schools (Municipal Educational Network or municipal administrative sphere); and (ii) States, which manage the state schools (State Educational Network or state administrative sphere). Since Resolution 26 establishes different amounts of federal transfers according to class characteristics (listed in Table 1, on page), FNDE is able to track the amounts transferred to each Municipality and State separately per transference category. Table 2 bellow, reports the distribution of total the federal transfers in 2013 per category. The year of reference for this database used in the study is 2013.

Table 2: FNDE Education Categories for federal transfers (US$ of 2013)

FNDE Educ. Categories Transfer (1000

R$) % Students (1000 stud.) % Primary School 2.006.948 56,7% 25.124 59,6% Secondary School 473.423 13,4% 7.506 17,8% Youth/Adults 196.412 5,5% 3.719 8,8% Preschool 392.903 11,1% 3.532 8,4% Nursery 391.312 11,1% 1.602 3,8% Indigenous 27.367 0,8% 236 0,6% Special School 24.198 0,7% 223 0,5% Quilombola 26.457 0,7% 201 0,5% Total 3.539.021 100,0% 42.142 100,0%

Source: FNDE (2013), School Census (2012).

(18)

17 c. Self-Reported Counterpart Expenses (FNDE, 2013)

Local governments (municipality or state) self-report to FNDE all their counterpart expenses. As described in section 2. Brazilian School Feeding Process, besides federal transfers, local governments also execute own resource expenses in order to provide school feeding. Municipalities report yearly their expenses in two categories: Food and Other expenses. These expenses will be called counterpart expenses throughout the rest of this study. The year of reference for this database used in the study is 2013.

d. Food Buying Invoices (FNDE, 2013)

FNDE has to guarantee that students enrolled in public schools receive the meals established in Resolution 26 in compliance with Law nº 11,947 that regulates adequate school feeding, as it is the main manager of PNAE. As a result, FNDE monitors how public resources allocated to PNAE are spent on purchasing food for meals preparation. This monitoring activity serves both as an accountability mechanism and a manner of verifying the nutritional properties of purchased good. Municipalities and Secretariats send all invoices regarding PNAE’s food purchases to FNDE, which are compiled and comprise this dataset. The year of reference for this database used in the study is 2013.

e. Human Development Atlas of Brazil (IPEA/UNDP, 2013)

This database is the result of a study carried out in partnership by the United Nations Development Program (UNPD) and the Institute for Applied Economic Research (IPEA/Brazil). It consists of detailed socio-demographic data about all 5.565 municipalities in Brazil. Variables describe city features such as demographics, education, income, employment, housing and vulnerability. The year of reference for this database used in the study is 2013.

f. Family Budget Survey (IBGE, 2014)

The Family Budget survey typically takes place every 5 years in Brazil. It is conducted by the federal Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística - IBGE). It gathers household consumption information, providing per household inhabitant detailed information. One of the main uses of this dataset is for updating official price indexes in Brazil. This is one of the sources of information we are going to use to value representative budgets14. The year of reference for this database used in the study is 2014.

g. Wholesale and Retail Business for price consultation

Table 3 bellow shows the sources used to extract information on wholesale and retail business prices used to estimate infra-structure and other costs.

14 Note: retail prices are commonly different from wholesale prices and we can somehow capture this difference by using the

gap among wholesale prices index and household prices index. Nevertheless, the Family Budget Surveys are still informative, especially when there's no other detailed price information by region. Listing this dataset in the methodology does not mean we will use it.. In addition, the future interviews with representative municipalities will ask for prices as well.

(19)

18

Table 3: Sources of Price Consultation

Source Link

Preço dos Combustíveis http://www.precodoscombustiveis.com.br/

Wolkswagen http://www.vw.com.br/pt/servicos/plano_de_manutencao.html Revista 'O Carreteiro' http://www.revistaocarreteiro.com.br/modules/revista.php?recid=1037

Custo de Reparo e Manutenção em Caminhões Canavieiros

https://uspdigital.usp.br/siicusp/cdOnlineTrabalhoVisualizarResumo? numeroInscricaoTrabalho=5421&numeroEdicao=16

Revista 'O Carreteiro' http://www.revistaocarreteiro.com.br/modules/revista.php?recid=908 Revista 'O Carreteiro' http://www.revistaocarreteiro.com.br/modules/revista.php?recid=148&edid=16 Liquigas http://www.liquigasabc.com.br/pagina/produtos

Revista 'Auto-Esporte'

http://revistaautoesporte.globo.com/Revista/Autoesporte/0,,EMI293633-10142,00-INMETRO+DIVULGA+CONSUMO+DE+MODELOS.html

INMETRO http://www.inmetro.gov.br/consumidor/pbe/veiculos_leves_2013.pdf CDA FIAT http://www.cdafiat.com.br/pdf/tabela-revisao-fiat.pdf

Casas Bahia http://www.casasbahia.com.br/

Pão de Açúcar http://www.paodeacucar.com.br/

Americanas http://www.americanas.com.br/

Magazine Luiza http://www.magazineluiza.com.br/

Leroy Merlin https://www.leroymerlin.com.br/

Casa Arco http://www.casaarco.com.br/ JM Atacado

http://www.jmatacado.com.br/loja/products/KIT-MERENDA-ESCOLAR-3-P%C7S-935.html

Lojas Clima http://www.lojasclima.com.br/bandeja-retangular-hotel-53x40x2-aluminio-abc/p

FIPE http://www.fipe.org.br/ IBGE http://www.ibge.gov.br/home/

Mercado das Máquinas www.mercadomaquinas.com.br

Mercado Mineiro http://www.mercadomineiro.com.br/pesquisa/gas-cozinha-pesquisa-precos Guanabara 13 http://www.guanabara13.com.br

Zap Móveis http://www.zap.com.br/imoveis/fipe-zap-b/

Source: author elaboration. 3.2. Primary Data

Although two primary data collection processes were undertaken throughout this study, online survey and telephone interviews, only the last was used to calculate per student meal cost. A further detailing of the processes, as well as the reason for opting to carry out both types of data collection is further detailed in section 6. Data Collection Process. In this section, only the primary data used to calculate per student cost is exposed.

a. Survey on Municipality and State Governments: Counterpart Funding (C-MICRO - FGV/WFP/FNDE, 2014)

This database is a result of a primary data collection undertaken by C-MICRO – FGV, planned and structured by C-MICRO - FGV, WFP and FNDE. The need for collecting this data was driven by the lack of details on the distribution of the counterpart Other expenses among the different cost types. This database was constructed with information gathered through a telephone interviews done with 13 State Secretariats of Education.

b. Representative School Feeding Cost Structure (C-MICRO-FGV/WFP/FNDE, 2015)

This database was assembled from May to July, 2015, through primary data collection by C-MICRO - FGV. The main purpose for collecting this data is to determine representative school cost structures for different school/municipality/state profiles in Brazil. This information was gathered using telephone interviews in 39 representative schools.

(20)

19

4. Expected Outcome

As previously stated, this study is expected to provide details on the total cost of the PNAE implementation. The National Fund for Education Development (FNDE) provides data on the annual financing of the food component of the estimated total cost. However, there are some other costs related to the program implementation that are not organized or available, such as expenditure on infra-structure. In order to determine the total cost of the program (at state and municipal level) both components are expected to be measured. This cost estimation is necessary for the program cost-effectiveness estimation, to be carried out subsequently in another study, where one of the benefits is agricultural development.

With this in mind, the questions this study seeks to answer are:

 What are the federal costs of the Program?

 How much of the Program total cost does the federal government subsidize, besides their regular transfers for school food acquisition?

 What kind of cost decomposition of PNAE – in terms of inputs, outputs and processes – would be beneficial to other governments seeking to implement similar programs?

To answer these questions, the following activities were carried out:

 Determine other relevant data sources (local and state level) to calculate the cost;

 Propose possible methodologies to estimate the cost of PNAE;

 Identify direct and indirect costs of PNAE;

 identify the additional cost elements that should be included in the total cost of the Program, such as transportation, staff, kitchen utensils, and other operating costs;

 Investigate which are the transport and logistics costs (depending on the operating model);

 Investigate which are the operational costs - the costs of non-food products and direct services.

(21)

20

5. Methodology

This section presents a strategy to calculate the average cost per meal of a primary school in Brazil for the year of 2013. The average cost per meal may be represented by the following equation15:

𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑀𝑒𝑎𝑙 𝐶𝑜𝑠𝑡 =𝑇𝑜𝑡𝑎𝑙 𝑀𝑒𝑎𝑙 𝐶𝑜𝑠𝑡

𝑀𝑒𝑎𝑙𝑠 𝑆𝑒𝑟𝑣𝑒𝑑 (1)

Total Meal Cost can be divided into two parts, Total Food Costs and Infrastructure and Other Costs, so the above equation may be rewritten as:

𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑀𝑒𝑎𝑙 𝐶𝑜𝑠𝑡 =𝑇𝑜𝑡𝑎𝑙 𝐹𝑜𝑜𝑑 𝐶𝑜𝑠𝑡𝑠 + 𝐼𝑛𝑓𝑟𝑎 𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒 & 𝑂𝑡ℎ𝑒𝑟 𝐶𝑜𝑠𝑡𝑠

𝑀𝑒𝑎𝑙𝑠 𝑆𝑒𝑟𝑣𝑒𝑑 (2)

This section is divided into three separate subsections related to each term of equation (2), namely:

5.1. Meals Served 5.2. Total Food Costs

5.3. Infra-Structure & Other Costs

It is important to note that the methodology adopted to calculate school feeding expenses considers the government point-of-view (the provider) and is not based solely on the amount disbursed. In other words, school feeding incurred expenses might vary according to student dropout rate, while government transfers associated to PNAE (official budget) are based solely on the total number of eligible students according to the School Census of the previous year. Therefore, incurred and transferred cost (from the demand and supply sides, respectively) may differ due to the fact that Federal Transfers for school feeding depend only on the number of eligible students, while the schools have food costs proportional to the number of attending students16 (including transferred students and dropouts).

The average cost per meal is also estimated for the 27 Brazilian states, i.e. a cost per meal for each region and state are estimated, besides a national average. The results illustrate the existing cost heterogeneity in Brazil, when presented by region and socioeconomic condition.

5.1. Meals Served

This section presents the strategy used to determine the number of total meals served. To obtain this quantity, the number of daily meals students may receive as established under Resolution 26 is used. Note that the number of daily meals varies according to student educational level. For instance, Primary Education students may receive 1 to 3 meals per day, proportionally to how many hours they regularly spend in school (presented in Table 4 below).

15 Using only quantities from Primary Education.

16 The schools usually calculate how much food of the total monthly amount provided by the municipality, they will use on a daily basis according to the number of attending students.

(22)

21

Table 4: Minimum of meals and Nutritional Needs of those meals that need to be offered to Public School Students per School Day, among FNDE Categories

FNDE Categories Time on

school17

Minimum of meals

Nutritional needs

Nursery Part time 2 30%

Indigenous or Quilombola Education - 2 30% Basic Education* Part time 1 (at most) 20%

Basic Education Part time 2 30%

"Mais Educação" Educational Program* Full Time 3 70%

Full time* Full Time 3 70%

*: Categories in which Primary Education Students are included.

Source: Resolution 26, 2013.

As such, the total number of Meals Served in a year may be calculated as shown in equation (3) bellow.

𝑀𝑒𝑎𝑙𝑠 𝑆𝑒𝑟𝑣𝑒𝑑 = [𝑆𝑡𝑢𝑑. 𝐹𝑢𝑙𝑙 𝑇𝑖𝑚𝑒 × 3 𝑀𝑒𝑎𝑙𝑠 + 𝑆𝑡𝑢𝑑. 𝑃𝑎𝑟𝑡 𝑇𝑖𝑚𝑒 × 1 𝑀𝑒𝑎𝑙]

× 𝑆𝑐ℎ𝑜𝑜𝑙 𝐷𝑎𝑦𝑠 (3)

There are 200 School Days in a calendar year (established under Resolution 26). The School Census provides information on the number of students that regularly stay up to 4 hours at school (Part Time Students) as well as the number of students who spend more than 4 hours at school (Full Time Students). With this, the total amount of meals served is obtained using secondary data on the number of part and full time students directly inserted into equation (3) above.

5.2. Total Food Costs

This section presents the strategy for calculating total food costs. As presented in Figure 2 on page 12, both Federal and Local Transfers (hereon referred to as Counterpart Transfers) finance Food Costs. Mathematically, this may be expressed as in equation (4) bellow:

𝑇𝑜𝑡𝑎𝑙 𝐹𝑜𝑜𝑑 𝐶𝑜𝑠𝑡𝑠 = 𝐹𝑒𝑑𝑒𝑟𝑎𝑙 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠 + 𝐿𝑜𝑐𝑎𝑙 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠 𝑓𝑜𝑟 𝐹𝑜𝑜𝑑 (4)

5.2.1. Federal Transfers

Resolution 26 establishes that Federal Transfers must be spent only on food. The dataset of the Official Registration of PNAE Federal Funding contains this information and may be classified according to the categories shown in Table 2 (on page 15). Transfers received in 2013 by public Primary Schools equals a total sum of R$2.006.948.000 for spending on food acquisition by public primary schools for their students. In using this number, two assumptions are made:

Assumption 1 “Primary School” category reported as in Table 1 contains all Brazilian

students enrolled in Primary Education.

We will omit primary education students enrolled in categories other than “Primary School” reported as in Table 1, namely “Youth/Adult Schools”, “Indigenous”,

(23)

22 “Special School” and “Quilombola” 18 students. The main reasons for dropping these categories from the analysis are:

(a) the share of primary school students enrolled in these categories represents only a small share of the total amount of students enrolled in Public Primary education in Brazil, therefore results will not suffer a statistically significant shift incorporating these groups into the dataset;

(b) the aforementioned groups present particular socioeconomic characteristics, which make them not comparable to average primary education students in Brazil, which in turn creates a larger barrier to extrapolate analysis to average primary education students around the world.

Assumption 2 Schools under State/Local Administrative Spheres contain all Brazilian

students enrolled in Public Primary Education.

The FNDE database only reports Federal Transfers paid to Municipalities and State Secretaries of Education. Therefore, it is not possible to account for federal transfers to students enrolled in primary education in Federal Public Schools. Given that most Federal Schools do not offer Primary Education, the amount of transfers left out is assumed to be negligible (as shown in Table 5 below).

Table 5: Distribution of Enrolled Students among School Administrative Spheres

Administrative Sphere Students % Municipal 25,896,349 47.29% State 20,097,307 36.70% Private 8,484,870 15.50% Federal 278,580 0.51% Total 54,757,106 100.00%

Source: School Census (2012).

From here on, all results will consider only schools of the Local Administrative Spheres and Private Schools in agreement with local government (State and Municipality)19.

5.2.2. Local Transfers for Food (LTF)

The amount of counterpart resources received by public schools may be split into “food” and “other expenses” for each city. The data for LFT is obtained using the database of Self-Reported Counterpart Funding, where Total Local Transfers for Food is simply the total amount spent with food for each city. However, differently from the Federal Transfers database, the Local Transfers database only contains the aggregated value of Local Transfers; it does not separate expenditure per FNDE Education Category (as shown in Table 2 on page 15). The strategy adopted approximated the number of local resources that go to primary school by weighing the total spent with local transfers by the number of students in primary education, as shown in equation (5) bellow.

LTF = Total LTF x Weighted Proportion of Students in Primary Education

(5)

18 Quilombola refers to rural african-brasilian communities formed by slave descendants. More information can be found at: http://basilio.fundaj.gov.br/.

19 There may be cases where private schools offer public education in the absence of public schools. In these cases, they are financed by public transfers.

(24)

23 Where: 𝑊𝑒𝑖𝑔ℎ𝑡𝑒𝑑 𝑃𝑟𝑜𝑝𝑜𝑟𝑡𝑖𝑜𝑛 𝑜𝑓 𝑆𝑡𝑢𝑑𝑒𝑛𝑡𝑠 𝑜𝑛 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 = 𝑆𝑡𝑢𝑑𝑒𝑛𝑡𝑠 𝑜𝑛 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 × 𝑀𝑒𝑎𝑙𝑠 𝑆𝑒𝑟𝑣𝑒𝑑 𝑜𝑛 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 ∑(𝑆𝑡𝑢𝑑𝑒𝑛𝑡𝑠 𝑜𝑛 𝐿𝑒𝑣𝑒𝑙 𝑜𝑓 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛𝑖 × 𝑀𝑒𝑎𝑙𝑠 𝑆𝑒𝑟𝑣𝑒 𝑜𝑛 𝐿𝑒𝑣𝑒𝑙 𝑜𝑓 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛𝑖)

An illustration of the weights used in equation (5) is represented in

Table 6 below considering the numbers of students in Brazil as the numbers of a hypothetical city. Keep in mind that the number of meals is determined exogenously as shown in Table 4 (page 20).

Table 6: Illustration of the Weighted Proportion of Students on Primary Education of a Hypothetical City20

Category Students % Meals % Students x

Meals % Primary School 25.124 59,6% 1 8,3% 25.124 52,7% Secondary School 7.506 17,8% 1 8,3% 7.506 15,7% Youth/Adults 3.719 8,8% 1 8,3% 3.719 7,8% Preschool 3.532 8,4% 2 16,7% 7.064 14,8% Nursery 1.602 3,8% 2 16,7% 3.203 6,7% Indigenous 236 0,6% 2 16,7% 472 1,0% Special School 223 0,5% 1 8,3% 223 0,5% Quilombola 201 0,5% 2 16,7% 402 0,8% Total 42.142 100,0% 12 100,0% 47.713 100,0%

Source: Elaborated by the authors

Each city has its own weight, by number of students per educational level (in Table 2, page 15). It is noteworthy that the number of students is weighted by the number of meals they receive. If the proportion of students enrolled in primary education was the only measure used, it would impose that each student receives the same number of meals at school, which is not true (Table 4, page 20).

Equation (Erro! Fonte de referência não encontrada.) is calculated for each representative city then aggregated to obtain the national amount of Local Transfers for Food in Primary Schools (LTF). By doing this, the following assumption is implicitly made:

Assumption 3 Food Counterpart resources are distributed proportionally to the number

of students enrolled in each FNDE Category of Education, weighted by the total number meals students consume in the school.

It is expected that food counterpart transfers of a city is distributed more intensely in schools with more students and/or in schools where nutritional needs are higher. For this reason, the study uses this weighted proportion and assumes that food counterpart resources follow this characteristic. The weighted proportion can be calculated by creating a general unit of measure of “meals needed in Primary Schools” and by observing how it is distributed among different FNDE Education Categories. For this the distribution of students is taken from School Census and the total number of meals according to type of school and student is obtained using the rule laid out in Table 4 (page 20).

20For the sake of simplicity of this example, we consider that all students are part-time students. If not, the number of meals

(25)

24 In addition, given the database of counterpart funding consists of self-reported values by the Municipalities, the robustness of these numbers is verified by using the Food Buying Invoices dataset. The total spent according to Invoices subtracted from the amount of Federal Transfers is considered to be the total food expenses paid for by the local government.

It is important to note that the total amount of Food Buying Invoices could be used as the Total Food Cost. However, to yield more robust results different datasets from different sources were used, including the Official Register of PNAE Federal Funding that is highly trustworthy.

5.3. Infra-Structure & Other Costs

This section presents the methodology used to estimate the costs related to infrastructure and other costs21. As mentioned before, there is no data available on costs related to school resources and Counterpart Transfers do not contain a discriminated expenditure classification per cost category. For this reason, this is the most critical part of the study. For implementing this approach, 45 representative school cost profiles were selected22.

5.3.1. Defining Representative School Costs Profiles

Brazil is one of the largest countries in the world in terms of geographical extension and presents high socioeconomic heterogeneity among its geographic regions.

The cost structure for Infra-Structure & Other Costs most likely depends on the geographic region where the school is located. A school in the Northern region of Brazil may face a different logistic structure when compared to a school located in the Southeastern region. Other costs may also vary according to the geographic region due to other regional particularities, such as local customs (e.g. eating habits), weather, among others.

A second dimension that may significantly influence the cost structure is the city Socioeconomic Condition. In a city with relatively higher socioeconomic condition, kitchen assistants, cooks and other employees most likely earn more than those in a city with lower socioeconomic conditions. This may also be true for cooking expenses such as kitchen gas, electricity, etc.

A final important characteristic that may be decisive in defining the cost structure related to school feeding is school size. The returns to scale of the school feeding process may vary according to its number of students. For instance, while 3 employees in the kitchen are enough to serve food for 50 students, to feed 100 students no more than 4 employees may be needed. In other words, the cost with employee per student may not present a linear proportional increase as the number of students increase.

For these reasons and due to budget constraints, 45 school cost structures were established to represent the majority of the existing school cost structures in Brazil. The study combines

21 Other Costs include all other expenses related to the provision of food at primary schools, other than food purchase and infrastructure, such as electricity and water spent in making meals.

22 It is worth mentioning that the choice of 45 school profiles is the largest amount of data the study was able to obtain due to budget constraints. However, this number is able to adequately capture Brazilian socioeconomic heterogeneity and the diversity of local market structures. The research team collected primary data for assembling this dataset.

(26)

25 features of each of the dimensions described above in order to have these 45 representative cost structure profiles: Geographic Region, City Socioeconomic Condition and School Size. The total number of representative school profiles (forty five) is obtained by multiplying the characteristics described above as shown in Figure 4 below: 5 Geographic Regions, 3 Levels of City Socioeconomic Condition, and 3 School Sizes.

Figure 4: Three Determinant Dimensions in Brazilian School Feeding Cost Structure and its features

Source: Author elaboration, Resolution 26, 2013.

The study defined 45 clusters of school profiles so that any school in Brazil can be classified under one of the categories. For classifying a school in one of the presented profiles, the following definitions are adopted:

Geographic Region: Geopolitical Boundaries;

City Socioeconomic Condition: the cities are grouped as Low, Medium and High

Socioeconomic Conditions by their position among all Brazilian cities according to their Human Development Index (HDI). The 33rd and 67th percentiles of the HDI distribution are used as thresholds for determining under which category the city will be positioned: Low - bellow the 33rd percentile; Medium – between the 33rd and 67th percentile; High – above the 67th percentile.

School Size: Using data from the School Census on School Size, schools are classified

as Small, Average and Big. Once again, the position of the school in the Brazilian school size distribution determines under which category it will be placed, using the 33rd and 67th percentiles as thresholds: Small - bellow the 33rd percentile; Average – between the 33rd and 67th percentile; Big – above the 67th percentile.

Geographic Region

North

Northeast

Midwest

Southeast

South

City Socioeconomic Condition

Low

Medium

High

School Size

(27)

26 After calculating the aforementioned thresholds and classifying Brazilian schools among them, several interviews took place with municipalities/secretariats23 responsible for schools as well as schools. The interviews assisted in defining a cost-structure for each profile, from which the cost-structure for other schools will be extrapolated. Further details of the primary data collection are contained in the sections below.

5.3.2. Defining an Input-Usage Function for Each Representative Profile

Once the groups of the Representative Profiles were defined, a school feeding input-usage function for each of them was drawn from gathered information. 45 representative lists of input quantities used in the process of school feeding was drawn based on collected primary data regarding the quantities of inputs used in the administrative and operational processes of school feeding in every representative school. The inputs are listed in Table 7 below.

Table 7: School Feeding Inputs and Their Units of Measure Costs Type24 Potential measure

Cooking Gas m³/time Electricity kw/time

Logistics Expenses transportation and km Employees uniform Quantity

Payroll quantity / hierarchy

Employees training quantity of employees trained

Facilities m²

Kitchenware list of utensils

Appliances list of refrigerator, freezer and stove

Other List

Source: Own Elaboration with FNDE experts.

Three complementary sources of information (secondary data described in the Data Collection Process section) were used to obtain the quantity of inputs used in each of the 45 school cost structures, besides the two following instruments of primary data collection:

Representative School Feeding Cost Structure – Interview with Nutritionists / Managers: Several interviews took place with Nutritionists and/or Managers that

work in municipalities in the 15 City Groups25 in order to collect data on the 45 school profiles. The interviewed are municipalities and state secretaries pointed out by FNDE experts together with FGV to: (a) best represent the profiles; (b) be the ones most likely to have the best management and organization in each City Group, most likely to provide information with the best quality and precision regarding expenditures with school feeding.

Survey on Municipality Counterpart Funding: This survey assists in understanding

how Counterpart resources are spent on Infra-Structure & Other Costs. This dataset

23 Henceforth, secretariat is to be read as municipal secretariat of education.

24All School and Secretariats assets used in the School Feeding Process were priced with market prices consultations and had

their value corrected accordingly to Legal Depreciation Rules.

25 The number 15 is equal to 5 Geographic Regions times 3 City Socioeconomic Condition. These two dimensions define the city

(28)

27 refers to other costs indirectly incurred in throughout the process of meal provision at public primary schools.

5.3.3. Pricing Inputs of the Representative Input-Usage Function

Once Input-Usage functions for the 45 profiles are drawn, the inputs are priced. Prices vary within the 15 City Groups and, therefore, must be city specific. Information on the prices was collected using the following sources of information:

Representative School Feeding Cost Structure – Interview with Nutritionists / Managers: When asking about inputs, prices of the inputs were also inquired;

Survey on Municipality Counterpart Funding: The total amount spent in each cost

category is contained in the survey. By having the 45 Input-Usage functions, it is possible to divide the amount spent by the quantity of inputs used and therefore obtain an approximation of the prices of inputs per city;

Household Survey: It is possible to extract from this survey several local prices and

wages based on household consumption and earnings, which are similar to the ones incurred on by schools;

Several other sources of Price Information: Price indexes, Rent Indexes, Retail and Wholesale Business, among others (detailed in Section 3.1 – Secondary Databases). 5.3.4. Estimating Infra-Structure & Other Costs for each Brazilian School

After pricing each input of the 45 representative functions, the values are added up for every profile to obtain the total expenses of Infra-Structure & Other Costs for each of the 45 profiles. After this process in done, 45 priced Input-Usage functions are drawn and then extrapolated for all schools in Brazil. Finally, total estimated expenses for each Primary School are obtained and aggregated to obtain the Total Infra-Structure & Other Costs in Brazil.

5.4. Estimating Total Meal Cost per School

This section presents the two specifications adopted to estimate total cost per meal using a linear regression strategy. The details for the two specifications adopted are shown in table 8 bellow.

Table 8: Econometric Specifications Adopted for Estimating Infra-Structure and Other Costs

Specification Explanatory Variables

1 Regional dummies; HDI; squared HDI; school size; squared school size. 2 HDI; squared HDI; school size; squared school size; school characteristics

(access to water, sewage and electricity, school material - masonry).

Under specification 1, the Total Expenses of the “Primary School i” depends on the geographic region of the School, on HDI of the City where the school is located, the number of students (School Size) in the school, and squared city HDI26 and number of students27

26 Proxy for measuring marginal increase or decrease in infra-structure and other costs given the initial level of HDI the municipality possess.

(29)

28 squared (Proxy for measuring School Scale). The regression model to be estimated is represented below:

𝐸𝑥𝑝𝑒𝑛𝑠𝑒𝑠𝑖 = 𝛽0+ 𝛽1𝑁𝑜𝑟𝑡ℎ𝑖+ 𝛽2𝑁𝑜𝑟𝑡ℎ𝑒𝑎𝑠𝑡𝑖+ 𝛽3𝑆𝑜𝑢𝑡ℎ𝑖+ 𝛽4𝑆𝑜𝑢𝑡ℎ𝑒𝑎𝑠𝑡𝑖

+ 𝛽5𝐻𝐷𝐼𝑖+ 𝛽6(𝐻𝐷𝐼𝑖)² + 𝛽7𝑆𝑐ℎ𝑜𝑜𝑙𝑆𝑖𝑧𝑒𝑖+ 𝛽8(𝑆𝑐ℎ𝑜𝑜𝑙𝑆𝑖𝑧𝑒𝑖)2

+ 𝜀𝑖

Where:

 𝐸𝑥𝑝𝑒𝑛𝑠𝑒𝑠𝑖 is the amount of total Expenses of “Primary School i” (in R$ of 2013);

 𝑁𝑜𝑟𝑡ℎ𝑖 is the binary variable that indicates 1 if the “Primary School i” locates in the Region North and 0 otherwise;

 𝑁𝑜𝑟𝑡ℎ𝑒𝑎𝑠𝑡𝑖 is the binary variable that indicates 1 if the “Primary School i” locates in the Region Northeast and 0 otherwise;

 𝑆𝑜𝑢𝑡ℎ𝑖 is the binary variable that indicates 1 if the “Primary School i” locates in the Region South and 0 otherwise;

 𝑆𝑜𝑢𝑡ℎ𝑒𝑎𝑠𝑡𝑖 is the binary variable that indicates 1 if the “Primary School i” locates in the Region Southeast and 0 otherwise;

 𝐻𝐷𝐼𝑖 is the HDI score of the city where the “Primary School i” is located;

 (𝐻𝐷𝐼𝑖)² is the HDI score squared of the city where the “Primary School i” is located;

 𝑆𝑐ℎ𝑜𝑜𝑙𝑆𝑖𝑧𝑒𝑖 is the number of students enrolled in “Primary School i”;

 (𝑆𝑐ℎ𝑜𝑜𝑙𝑆𝑖𝑧𝑒𝑖)2 is the squared number of students enrolled in “Primary School i”; 𝜀𝑖 is the model error term related to the “Primary School i”.

(6)

Analogously to the explanation of specification 1, equation (7) shows specification 2 (of Table 8):

Expenses

i

0

+ β

1

HDI

i

+ β

2

(HDI

i

)

2

+ β

3

SchoolSize

i

+ β

4

(SchoolSize

i

)

2

+

β

5

Water

i

+ β

6

Waste

i

+ β

7

Sewage

i

+ β

8

Electricity

i

+ β

9

Masonry

i

+ ε

i

(7)

Where:

Expensesi is the amount of total Expenses of “Primary School i” (in R$ of 2013);

HDIi is HDI score of the city where the “Primary School i” is located;

(HDIi)2 is HDI score squared of the city where the “Primary School i” is located;𝑆𝑐ℎ𝑜𝑜𝑙𝑆𝑖𝑧𝑒𝑖 is the students quantity enrolled on the “Primary School i”;

SchoolSizei is the number of students enrolled in the “Primary School i”;

(SchoolSizei)2 is the squared number of students enrolled in the “Primary School i”;

Wateri is the number of households with piped water and bathroom in the municipality where “Primary School i” is located;

Wastei is the number of households with access to adequate waste collection system in the municipality where “Primary School i” is located;

Sewagei is the number of households with inadequate sewage treatment in the municipality where “Primary School i” is located;

Electricityi is the number of households with access to electricity in the municipality where “Primary School i” is located;

 Masonryi is the number of households without masonry walls in the municipality where “Primary School i” is located;

 εi is the model error term related to the “Primary School i”.

By estimating the interest parameters of this model with the 45 observations from the interviews with Secretariats/Municipalities Nutritionists and Managers, one is able to predict the cost per meal for each Brazilian primary school. This extrapolation process is done by using the fitted regression line and the explanatory variable values of each school present in the School Census Dataset.

(30)

29

6. Data Collection Process

This section presents the two primary data collection processes undertaken throughout the cost per meal assessment. It is comprised of two subsection: 6.2. Online Survey and 6.3. Interviews: Telephone and In-person, each describe two separate data collection processes undertaken by C-micro for the execution of this project. The need for primary data collection was a result of the lack of detailed expense description on behalf of the counterpart (municipalities and schools), especially regarding infra-structure, utensils, equipment and other inputs.

6.1. Online Survey

This section contains the main results found with the online survey, subdivided into three section: 6.1.1. Answers from Secretariats; and 6.1.2. Answers from Municipalities, highlighting the main challenges faced in obtaining reliable information with this tool.

Initially, an online survey was proposed in November/December of 2014, were the potential respondents were all managers and nutritionists that work directly with PNAE, in each Municipality and State Secretary of Education. They were asked to discriminate the total amount of yearly self-reported counterpart expenses value in 11 cost categories, as shown in Table 9 bellow.

Table 9: Cost Categories on which Counterpart Budget was Spent Counterpart Budget 1. Food buying 2. Storage 3. Cooking Gas 4. Electricity 5. Logistics Expenses 6. Employees uniform 7. Payroll 8. Employees training 9. Infra-structure 10. Kitchenware 11. Other

Source: FNDE, C-Micro, WFP (2014).

Survey inquiries were sent to 12,369 workers inside each Municipality and State Secretary of Education, among them nutritionists and managers. Only 3,571 of the total workers opened the online survey and started to answer it. Of this number, 1,030 completed the survey, however only 652 were informative enough to address the lack of data regarding distribution of counterpart resources.

Furthermore, as the information derives from self-reported surveys, i.e. not official, additional care is needed throughout the data mining process. A cross-check was made using the Self-Reported Counterpart Database from the FNDE (2013). Only 151 completed answers were consistent with this dataset. The steps of this data mining process are listed in Table 10 on following page.

The subsections bellow describes the results obtained with the online survey and main challenges of using this information to calculate the average school feeding cost per meal.

Imagem

Figure 1: Dimensions of a School Class and Eligibility for Public Funding for Food (categories marked in red are  not eligible to PNAE)
Table 1: Daily per capita federal transfer for a student according to class type (US$ of 2013) 9
Figure 3: Inputs for PNAE and the Origin of the Funds Used
Table 2 bellow, reports the distribution of total the federal transfers in 2013 per category
+7

Referências

Documentos relacionados

In addition, decompositions of the racial hourly earnings gap in 2010 using the aggregated classification provided by the 2010 Census with 8 groups, instead of the one adopted in

Controls: Year dummies, school size (number of students), average teacher years of schooling, student- to-teacher ratio, number of students per class, dummy for sewage at the

The probability of attending school four our group of interest in this region increased by 6.5 percentage points after the expansion of the Bolsa Família program in 2007 and

No campo, os efeitos da seca e da privatiza- ção dos recursos recaíram principalmente sobre agricultores familiares, que mobilizaram as comunidades rurais organizadas e as agências

Peso de nódulos por vaso, em duas cultivares de caupi que receberam inóculo de quatro estirpes de Bradyrhizobium spp, em função dos níveis de calagem.. " Não foram usados

The mean number of hatchlings per nest was determined from the sum of the number of live and dead hatchlings divided by the total number of nests, while the mean number of eggs

Thus, the 2008 impact factor, for instance, was given by the number of papers published in this journal in 2006 and 2007 that were cited by ISI indexed journals divided by the number

The average weight of eggs was multiplied by the total number of eggs produced during the experimental period, obtaining the total eggs mass, which was divided by the total number