Comissão Parlamentar de Inquérito ao Pagamento de Rendas ...
Ensaios Econômicos - cps.fgv.br · indicadores sociais, como renda total per capita, ganhos...
Transcript of Ensaios Econômicos - cps.fgv.br · indicadores sociais, como renda total per capita, ganhos...
Ensaios Econômicos
Escola de
Pós-Graduação
em Economia
da Fundação
Getulio Vargas
N◦ 792 ISSN 0104-8910
Infrastructure investment and social progressin Brazil
Marcelo Neri
Dezembro de 2017
URL: http://hdl.handle.net/10438/19456
Os artigos publicados são de inteira responsabilidade de seus autores. Asopiniões neles emitidas não exprimem, necessariamente, o ponto de vista daFundação Getulio Vargas.
ESCOLA DE PÓS-GRADUAÇÃO EM ECONOMIA
Diretor Geral: Rubens Penha CysneVice-Diretor: Aloisio AraujoDiretor de Ensino: Caio AlmeidaDiretor de Pesquisa: Humberto MoreiraVice-Diretores de Graduação: André Arruda Villela & Luis Henrique Bertolino Braido
Neri, MarceloInfrastructure investment and social progress in Brazil/
Marcelo Neri – Rio de Janeiro : FGV,EPGE, 2017107p. - (Ensaios Econômicos; 792)
Inclui bibliografia.
CDD-330
1
Infrastructure Investment
and Social Progress in
Brazil1
Marcelo Neri2
FGV Social and EPGE/FGV
Abstract
This paper draws a broad empirical diagnosis on the evolution of infrastructure coverage
in Brazil and potential social impacts. It focuses on the sectors of sewerage, water,
electricity, urban transportation and information and communication technologies (ICTs).
Most of the analysis departs from household surveys, bringing the population perspective
into the picture. We analyze socio-economic determinants of infrastructure coverage, a
social outcome in itself, as well as their possible indirect impacts on income generation,
time cost of transportation, housing values and education. We also consider briefly direct
consequences of increasing infrastructure coverage in the budget constraint through
services costs and payments delays and direct utility effects through subjective data on
the quality and importance attributed to different infrastructure sectors.
1 This paper was prepared as a background paper for a World Bank project on Infrastructure in Brazil. It extends a series of previous researches carried out by FGV Social for the World Bank and also benefits from previous work performed in sewerage and water for the NGO Trata Brasil, on ICTS for Telefonica Company and on Urban Transportation for the State of Rio de Janeiro. (see WWW.fv.br/cps/tratabrasil5). I would like to thank the excellent research assistance provided by Luisa Melo, Samanta Sacramento, Manuel Osorio and Thiago Cavalcante. I would also like to thank the comments provided in earlier versions of the paper by Edith Kikoni and Marcos Hecksher. 2 [email protected]
2
In all stages of the analysis we look for causes and consequences of coverage, static and
dynamic, bivariate and multivariate, emphasizing the roles played by two variables:
income and the geographical dimension in order to locate necessary public policies efforts
in the social strata and in the territory. For example, multivariate analysis allows
comparing access of individuals with the same observable characteristics (income, city
size etc.) across different units of federation. This allows us to map repressed demand for
a future infrastructure expansion. The income dimension enters both as a determinant as
well as a consequence of infrastructure coverage. We test how much “exogenous”
household income increases related to expansion of Bolsa Família as an experiment of
pure income effects channels on infrastructure outcomes. The reverse channel studies
how much infrastructure affects the income convergence across municipalities and of
other social dimensions such as poverty, human development and its components. It
explores the convergence issue in a more general setting through quantile regressions.
We estimate the potential impacts of different infrastructure sectors along the distribution
of various social indicators such as per capita total income, labor earnings, education,
imputed rents, commuting costs and a constructed broader social welfare measure. We
also apply a variable selection procedure to rank infrastructure variables in terms of
potential social impact on poverty and those indicators discussed above. This exercise
includes externality effects of infrastructure at the community level on individual social
outcomes, a market failure that may justify as well as signal the necessity of certain policy
interventions. We end analyzing potential infrastructure impacts on school flows and
proficiency from SAEB microdata.
Resumo
Este artigo desenha um diagnóstico empírico amplo sobre o nível e a evolução da
cobertura de infraestrutura no Brasil e seu potencial impacto social. Centra-se nos setores
de esgoto, água, eletricidade, transporte urbano e Tecnologias de Informação e
Comunicação (TICs). A maior parte da análise parte das pesquisas domiciliares, trazendo
a perspectiva da população sobre o tema. Analisamos os determinantes socioeconômicos
da cobertura da infraestrutura, um resultado social em si, bem como seus possíveis
impactos indiretos na geração de renda, no custo do tempo de transporte, nos valores da
habitação e na educação. Também consideramos as consequências diretas do aumento do
investimento em infraestrutura na restrição orçamentária através de custos de serviços e
3
atrasos de pagamentos e efeitos diretos na de utilidade através de dados subjetivos sobre
a qualidade e a importância atribuída a diferentes setores de infraestrutura.
Em todas as etapas da análise, buscamos causas e consequências da cobertura, estática e
dinâmica, bivariada e multivariada, enfatizando os papéis desempenhados por duas
variáveis: a renda per capita e a dimensão geográfica para localizar os esforços
necessários de políticas públicas nos estratos sociais e no território. Por exemplo, no caso
do nível de cobertura, a análise multivariada permite comparar o acesso de indivíduos
com as mesmas características observáveis (renda, tamanho da família, educação, gênero,
tamanho da cidade, etc.) em diferentes unidades de federação. Isso nos permite mapear a
demanda reprimida para uma futura expansão de infraestrutura. A dimensão da renda
entra tanto como determinante como também como uma consequência da cobertura da
infraestrutura. Testamos o quanto mudanças "exógenas" da renda relacionada à expansão
do Bolsa Família impactam os resultados da infraestrutura, como um experimento de
canais de efeitos de renda pura. O canal reverso estuda o quanto infraestrutura afeta a
convergência de renda entre os municípios e de outras dimensões sociais como pobreza,
desenvolvimento humano e seus componentes. Ele explora a questão da convergência em
uma configuração mais geral através de regressões quantílicas. Estimamos os impactos
potenciais de diferentes setores de infraestrutura ao longo da distribuição de vários
indicadores sociais, como renda total per capita, ganhos trabalhistas, educação, rendas
imputadas, custos de deslocamento e uma medida de bem-estar social mais amplamente
construída a partir de alguns destes elementos. Também aplicamos um procedimento de
seleção de variáveis para classificar variáveis de infraestrutura em termos de impacto
social potencial sobre a pobreza e os indicadores discutidos acima. Este exercício inclui
efeitos de externalidades emanadas pela infraestrutura a nível comunitário sobre os
resultados sociais individuais, uma falha do mercado que pode justificar e sinalizar a
necessidade de certas intervenções de políticas. Terminamos analisando possíveis
impactos de infraestrutura nos fluxos e proficiência escolares através de microdados do
SAEB.
4
Executive Summary
Empirical Diagnosis on Infrastructure Coverage
Conceptual Framework - The potential social impacts channels of infrastructure changes
will be assessed here under three headings. First, the direct impact on well-being: we will
initially interpret the data on coverage rate as a direct social impact in itself. This is the
main line of inquiry pursued here. Additionally, subjective data will add quality and
priority measures of different infrastructure coverages. Second, the direct impact on the
monetary budget constraint depending on the way infrastructure services are financed.
This evidence will be limited by the scarcity of more recent data. A third channel largely
explored here is the influence exerted by infrastructure on individual income and assets
generation process, modeled by movements along the production function, shifts in the
production function and on the way individuals connect to inputs and outputs markets –
for example, in the case of transportation and ICTs. We run income regressions exploring
the interaction of infrastructure with human capital and social economic characteristics.
Hedonic rent equations add evidence of the infrastructure effects on assets value since
housing is the most important physical asset. Similarly, we perform exercises for the time
opportunity cost of transportation and for broader societal well-being indicators that
include all above. Equations on the impact of infrastructure on years of schooling, grade
repetition and proficiency complement the analysis.
Access to infrastructure services has increased significantly over the past decade.
This is mainly due to lagged effects of the privatization programs of the 1990s (especially
in telecommunications), the adoption of public programs aimed at expanding coverage in
remote areas (especially in electricity due to the “Luz Para Todos” program) and the
demand effect from the combination of faster household income growth and falling
inequality that lasted until 2014. Using household level data on coverage of infrastructure
services, the service that had the highest increase in access between 2004 and 2015 was
ICT. The past 10 years has seen an explosion in the use of mobile telephones. In 2004,
around 85 million people had mobile phones at home, and in 2015 the number increased
to 186 million – an increase of 101 million users. During the same period, home internet
coverage was extended to additional 64 million Brazilians. Despite its rapid growth,
internet service is the infrastructure service that presents the lowest level of access (42.5
percent) when compared to other services. On the other extreme is electricity, with an
5
access level of 99.7 percent. Access to potable water has an intermediate rate of 83.6
percent, but significantly more than sewage services, at 56.9 percent.
Brazil: share of population with access to infrastructure services (%)
Source: FGV Social/CPS from PNAD/IBGE microdata
While there has been some convergence over the past decade, significant regional
differences remain across the country in terms of access to infrastructure services,
particularly in water, sewerage and internet services.
Coverage of infrastructure services in rural areas has expanded but the sharp divide
between rural and urban coverage within the country persists. Only in sewerage has
rural coverage not changed much. However, access gaps between rural and urban areas
remain high. While rural areas represent around 14 percent of the Brazilian population in
2015, only 4 percent of this population has access to sewage services with only a third
having access to the water system. In urban areas, where most of the population lives, the
rate of access to the water system is about 90 percent, while access to sewage services is
about 80 percent. The pattern of low rates of access in rural areas and high rates of access
in urban areas is evident in all infrastructure services with the exception of electricity,
where access rates have converged.
Infrastructure access reflects and reinforces Brazil’s high income inequality profile.
Access rates among the poor have been improving in the last decade but coverage remains
much higher among wealthier groups. Sewerage, water and internet tend to be the most
unequally distributed services across income groups. In 2015, less than half of the poorest
segment of the population had access to sanitation facilities, compared with 80 percent of
the richest.
11.5
45.1
77.4
47.8
96.3
42.5
56.9
83.693.5 99.7
0
20
40
60
80
100
internet sewerage water cellphone electricity2004 2015
6
Income Group – % Infrastructure Coverage
7
Type of Area – % Infrastructure Coverage
8
Macro-Region – % Infrastructure Coverage
9
Infrastructure Conditional Coverage Convergence Across States
Multivariate exercises – We ran regressions to isolate the determinants of infrastructure
coverage in the period of analysis. Besides gender, race and spatial variables, we use
second degree polynomials for per capita income, family size, education and age. These
quadratic terms turned out significant in most of the regressions.
Year effects - Keeping socio-demographic structure constant, the highest temporal change
between 2004 and 2015 was observed in electricity, internet and cell phone. The lowest
expansion was found in water and sewerage.
Family size effects were positive but at diminishing rates. This point is noteworthy since
as a product of the demographic transition household size had fallen 1.43% per year, a
path that was faster than the 0.8% per year of total population size growth rate. This means
that the infrastructure supply has to increase not only because of the existing infrastructure
deficit and population growth but also as a response to the household size reduction.
States Evolution – Many of the spatial differences of infrastructure coverage may be
attributed to differences in income, education, family size, city size, states and so on. We
focus our analysis on the later spatial variable. The maps presented in each page present
the geographical dispersion of coverage across Brazilian states. São Paulo is always
portrait white as the basis (i.e. the omitted variable). The red means that is lower than São
Paulo, while blue gives the excess with respect to São Paulo. As a general rule, São Paulo
presents the best infrastructure across States in the country.
Next we run an extension of the previous multivariate exercise also incorporating the
interaction between State Dummies and year in order to grasp the spatial dimension of
infrastructure coverage changes. We also fixed São Paulo as the omitted spatial dummy
and 2004 as the omitted temporal category. In this way the results are directly interpreted
as the conditional difference in difference of each state in 2015 with respect to São Paulo
in 2004. Or how much the infrastructure coverage changed in relative terms. In most cases
the color of the map turns into blue which means that the differential between different
states and São Paulo tended to fall. This shows a clear convergence trend of infrastructure
between Brazilian states even if we net out the effects of income, education and other
variables during this period.
10
Electricity – States Odds Ratio http://cps.fgv.br/razao_Has Access to Electricity
Electricity – States Odds Ratio with Time Interaction http://cps.fgv.br/razao_Has Access to Electricity_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
11
Water (General Network) – States Odds Ratio http://cps.fgv.br/razao_agua
Water (General Network) – States Odds Ratio with Time Interaction http://cps.fgv.br/razao_agua_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
12
Sewerage – States Odds Ratio http://cps.fgv.br/esgoto_razao
Sewerage – States Odds Ratio with Time Interaction http://cps.fgv.br/esgoto_razao_com_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
13
Computer with Internet at Home – States Odds Ratio
http://cps.fgv.br/razao_comp_com_net
Computer with Internet at Home – States Odds Ratio with Time Interaction http://cps.fgv.br/razao_comp_com_net_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
14
Landline or Mobile Phone at Home – States Odds Ratio
http://cps.fgv.br/razao_fixo_celular
Landline or Mobile Phone at Home – States Odds Ratio with Time Interaction http://cps.fgv.br/razao_fixo_celular_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
15
Ln Equation of Course Time Evaluated at Hourly-Wage (Main Job) – States Odds Ratio http://cps.fgv.br/estimativa_equacao_de_ln_de_tempo_de_percurso_ao_trab
Ln Equation of Course Time Evaluated at Hourly-Wage (Main Job) – States Odds Ratio with Time Interaction
http://cps.fgv.br/interacao_equacao_de_ln_de_tempo_de_percurso_ao_trab
Source: FGV Social/CPS from the PNAD/IBGE microdata
16
Causality and the Bolsa Família Experiment – Next, we use the expansion of Bolsa
Família to test the effect of “exogenous” income changes on access to public services.
The effect is captured by a difference-in-difference estimator generated from the
interaction of the dummy variable year (before and after expansion) with the dummy
variable for the program eligibility criterion (per capita household income less than
R$100 monthly in real terms, excluding income earned by social programs). How much
the increase of the access to public services is related to the increase of income of this
population through the expansion of 67% of the program coverage between 2004 and
2006. We used the dummy variables above (eligible*year) to measure whether the income
gain of the low-income population increased more than the others. The results are a
relative improvement for all items (except sanitation). In the case of cell phone access
and landline telephone the chances are 13% and 11% higher, while in access to public
services, such as garbage collection, electricity and general water network, the chances
are 13%, 11% and 8% higher, respectively. The improvement of transportation is captured
by a -1.3% fall in commuting time at individual level. The same goes for assets such as
computer connected to the internet and bathroom at home. However, for sewerage
connected to the general network there was no statistically significant improvement in
relation to the other group. Thus, the higher income did not impact access to the sewerage
network of the population eligible to the program. This lack of sensibility may be due to
the predominance of externalities in the supply of sewerage where individual or private
returns to sewerage connection benefits mostly others.
Perceived Quality and Priorities – The IBGE Household Budget Survey allows us to
explore the perceived quality of access to services. In general, the quality of services
associated with water enjoys lower perceived quality than that of public services such as
electricity and garbage collection. Besides the subjective quality attributed to each
infrastructure service, one may also investigate the weights given to them by the
population itself. An analysis of the priorities of the Brazilian population is made in terms
of public policy vis-à-vis the global population. Out of 16 new Sustainable Development
Goals (SDGs) related items, infrastructure variables stay in the following positions:
Transportation (7th); Water and Sanitation (9th); Electricity (13th) and ICTs (16th).
According to the global wide sample infrastructure priorities were: Water and Sanitation
(5th); Transportation (12th); ICTs (16th) and Electricity (15th).
17
Infrastructure and Social Convergence Across Cities
Standard Convergence – We followed initially the standard economic growth literature
and tested the role of infrastructure variables in terms of reducing inequality between
income and other social variables across 5500 Brazilian municipalities. We basically
implemented a standard convergence analysis running regressions of growth of each
variable against the natural logarithm of initial value comparing the results with and
without infrastructure variables. The set of variables tested includes per capita GDP, per
capita household income, the Human Development Index, its 3 components plus a series
of related variables such as poverty and inequality, life expectancy, child mortality, school
attendance for various age brackets and the Basic Education Development Index (IDEB)
which includes the results of proficiency exams. For 16 out of the 17 endogenous
variables tested, the speed of convergence is higher at face value with the set of
infrastructure variables than the model without infrastructure.
Source: FGV Social/CPS from the Demographic Census IBGE microdata; Ipea, UNDP and FJP (2013) and
INEP/MEC.
#This regression is for the endogenous variable in percentage against its variation in percentage points
##sample for 5010 cities between 2007 and 2015
Household income seems to capture better than GDP the infrastructure induced effects.
-0.253
-0.232
-0.043
-0.032
-0.040
-0.011
-0.149
-0.275
-0.069
-0.104
-0.094
-0.335
-0.151
-0.334
-0.154
-0.458
0.131
Basic Education Index (IDEB) for the 5th grade##
Basic Education Index (IDEB) for the 9th grade##
School Attendance - Children 0-3 years
School Attendance - Children 4-6 years
School Attendance - Children 6-14 years
School Attendance - Children 6-17 years
Life Expectancy
Gini Index
Poverty (Proportion of Poor)#
Child Mortality Under 1 year
Child Mortality Under 5 years
HDI Income Component
HDI Health Component
HDI Educational Component
Human Development Index (HDI)
Per capita Household Income
Per capita GDP
Regressions for Rates of Change across 5500 Municipalities between 2000-2010Difference Lagged Endogenous Variable Coefficient With and Without Infrastructure
18
Poverty rate regression was treated in levels with the results showed below.
Another statistics across these series of regressions that is worth looking at is the adjusted
R2. The gross explanatory power of infrastructure in terms of the various dimensions of
social changes ranges from 13.9% on child mortality to 66% for the Human Development
Index.
-1
-0.5
0
0.5
1
1.5
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
Esti
mat
ed p
er c
apit
a In
com
e V
aria
tio
n 2
00
0-2
01
0
LN (per capita Income in 2000)
Convergence in per capita Household Incomebetween Brazilian municipalities
Y estimated only with LN(pc Income) Y estimated w/ infrastructure variables
y = -0.264x - 6.991
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
End
oge
no
us
Var
iati
on
20
00
-10
(p
erce
nta
ge
po
ints
)
Endogenous 2000
Convergence
19
Source: FGV Social/CPS from the Demographic Census IBGE microdata; Ipea, UNDP and FJP (2013) and
INEP/MEC.
#this regression is for the endogenous variable in percentage against its variation in percentage points
##sample for 5010 cities between 2007 and 2015
The Human Development Index is a more encompassing measure of social progress.
Graph below illustrate the convergence of the Human Development Index.
Source: FGV Social/CPS from the Demographic Census IBGE microdata; Ipea, UNDP and FJP (2013)
The growth regression exercise with infrastructure variables as explanatory variables was
to some extend unsatisfactory, once the signs of the infrastructure variables were not
7.9%
3.5%
0.7%
8.6%
22.0%
20.2%
29.5%
5.1%
34.9%
13.9%
23.8%
28.0%
31.3%
60.3%
66.3%
15.6%
1.5%
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Basic Education Index (IDEB) for the 5th grade##
Basic Education Index (IDEB) for the 9th grade##
School Attendance - Children 0-3 years
School Attendance - Children 4-6 years
School Attendance - Children 6-14 years
School Attendance - Children 6-17 years
Life Expectancy
Gini Index
Poverty (Proportion of Poor)#
Child Mortality Under 1 year
Child Mortality Under 5 years
HDI Income Component
HDI Health Component
HDI Educational Component
Human Development Index (HDI)
Per capita Household Income
Per capita GDP
Regressions for Rates of Change across 5500 Municipalities between 2000-2010Gross Explanatory Power of 6 Infrastructure Variables R2
Gross Contribution
y = -1.514x - 0.154
-0.20
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
-0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00
End
oge
no
us
Var
iab
le V
aria
tio
n 2
00
0-1
0
LN (Endogenous) 2000
Convergence
20
robust. We should also test taking advantage of the availability of household surveys
microdata.
Before moving to the next step of the analysis it is useful to pose a few additional
questions, namely: Why income convergence between states in Brazil matters? There is a lot of
income inequality within states. Why not looking at overall inequality directly? Should we invest
in poor states, or in poor people anywhere in the country? Should we be looking in broader terms
a social welfare that combines lower overall inequality and higher overall growth?
Infrastructure and the Distribution of Social Outcomes
Quantiles Convergence – The next step is to construct a quantile regressions based
platform to test the social impacts, or at least the correlations between social outcomes
and the series of infrastructure variables. First, we construct from PNAD 2004 and 2015
microdata a series of social results variables which includes per capita household income
(total sources and labor earnings), years of schooling (for the whole population and for
people between 7 and 15 years of age). Imputed rents coming from a hedonic equation
and the opportunity time cost of commuting time evaluated at individual hourly wage
rates. We emphasize here the potential distributive impacts on the broader social welfare
measure (BSW) that includes total reported income plus imputed rent minus commuting
costs. We present here an analysis of various infrastructure items following an increasing
order of magnitudes around the median of BSW, starting with the lack of more traditional
public services and then access to ICTs, as shown in the graphs below.
Lack of Electricity – The coefficient by those who use oil, kerosene or gas as sources of
light in comparison with those that have electricity at home as a general rule presents a
robust negative sign in all results variables tested. In the case of our broader social welfare
(BSW) measure, coefficients are always negative and reach the bottom at the 60th
percentile. The distribution reaches -6.0% at the 40th percentile and -7.8% at the 90th
percentile. Lack of Water – The coefficient of those with no connection to water network
at home as a general rule also presents a robust negative sign in all results variables tested.
BSW coefficients are always negative and reach the least negative values around the
median. The distribution of coefficients reaches -20.0% at the 40th percentile and -19.2%
at the 90th percentile. Lack of Sewerage – Coefficients of those who live in dwellings
with rudimentary cesspit compared with those that have a sewerage network connection
21
at home presents a robust negative sign in all results variables tested, except years of
schooling for those at the age corresponding to primary level of education. BSW effect
increases almost monotonically in absolute value as we move to the upper tail of the
distribution, from -18.3% at the 40th percentile to -24.3% at the 90th percentile.
Lack of Public Services and Broader Social Welfare Measure Changes Across Vintiles
Source: FGV Social/CPS from the PNAD/IBGE microdata
Communication – We analyze the impact coefficient of those who are in dwellings with
telephone or cell phone for at least one of the household members compared to the rest
of the population without this device. Note that we are looking now for those who have
access compared with those who have not, so all the signs in the impact analysis of
infrastructure work the other way around. Most of the effect is due to cell phone
possession that became much more diffused than landline phone. As opposed to the
internet, the total income effect is higher than the labor earnings effect and both remain
higher than the rental value effect. The cell phone effect is relatively higher on the basis
of the distribution than internet access. The statistics organized by type of social outcome
show that: as a general rule, communication coefficients present a robust positive sign in
all results variables tested. BSW effects increases from 34.7% at the 40th percentile to
43.4% at the 90th percentile. The income variables related coefficients increase along each
particular concept. As a consequence, the diffusion of internet should lead to a divergence
in these different social outcomes. Internet – The impact coefficient of individuals in
-27
,3%
-26
,5% -24
,3%
-22
,6%
-22
,2%
-21
,8%
-21
,2%
-20
,0%
-19
,5%
-20
,0%
-19
,5%
-20
,1%
-19
,9%
-20
,0%
-19
,8%
-19
,1%
-19
,2%
-21
,7%
-19
,9% -1
6,7
%
-17
,0%
-17
,4%
-17
,6%
-17
,8%
-18
,1%
-18
,3%
-21
,9%
-23
,9%
-24
,3%
-25
,4%
-5,4
%
-5,6
%
-6,5
%
-6,3
%
-5,1
%
-5,9
%
-6,1
%
-6,0
%
-5,7
%
-6,8
%
-8,7
%
-10
,0% -8
,2%
-6,9
%
-7,0
%
-8,5
%
-7,5
%
-7,8
% -5,1
%
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95%
No Water Network No Sewarage No Electricity
22
dwellings with internet access compared with those without it presents a robust positive
and high sign in all results variables tested. The coefficients presents a positive trend as
we move towards the top of each distribution, suggesting at face value that those at the
top benefit relatively more from internet access. As a consequence, the diffusion of
internet should lead to a divergence in these different social outcomes. BSW effects
increases from 58.4% at the 40th percentile to 82.7% at the 90th percentile.
ICTs Coverage and Broader Social Welfare Measure Changes Across Vintiles
Source: FGV Social/CPS from the PNAD/IBGE microdata
Commuting time evaluated at hourly-wage rate – It works as an approximation to
transportation cost in urban areas and it is included in the broader welfare measure. We
just check whether it has increased from 2004 in 2015 and its distributive change pattern.
The 5% poorest had the highest increase of 41.1%, that tended to decrease, reaching
33.6% at the 40th percentile, with some stability reaching to 32.3% at the 90th percentile,
then rising to 35.4% at the top vintile.
Ranking Infrastructure Direct Social Impacts & their Externalities
Instead of imposing a particular model of analysis, we implement here a stepwise variable
selection procedure to determine which socio-economic and infrastructure related
variables are more statistically important to explain each social outcome variable seen
57
,2%
53
,8%
55
,1%
54
,9%
55
,6%
56
,3%
57
,0%
58
,4%
59
,3%
60
,0%
61
,5%
63
,3%
65
,5%
68
,2%
71
,0%
73
,9%
79
,0%
82
,7%
86
,6%
38
,6%
36
,2%
33
,8%
33
,1%
33
,6%
33
,8%
34
,1%
34
,7%
35
,5%
36
,0%
36
,9%
37
,3%
38
,0%
38
,8%
40
,0%
41
,2%
41
,9%
43
,4%
46
,2%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95%
Internet Phone
23
above. In the selection process we included externality effects from infrastructure. This
is done by including in the regressions the mean of these variables across geographic
areas. The idea is that beyond individual impacts at the household level, what other
community members have in terms of infrastructure may also affect our respective social
outcomes. For example, if there is a widespread diffusion of landline or cell phones in my
region of residence the value of my phone line increases due to network scales, given the
fixed cost of intercity connections.
Poverty – In the case of the proportion of the poor included at this stage, the six
infrastructure variables are significant in descending order: communication, internet,
transportation, water, electricity and sewerage. Two of the externality related variables
also presented statistically significant impacts, namely mean transportation time and
mean electricity coverage. Electricity access at the community level may improve
individual social outcomes through better work opportunities or school or health services.
Transportation use on the other extreme imply a common good congestion problem where
the excessive use of infrastructure generates a negative externality on all users.
Mean Broader Welfare – For broader social measure mean – that includes besides total
income sources from PNAD, imputed rents from housing minus opportunity time cost of
commuting – the results are similar to poverty. ICTs and transportation time present the
highest significance. Externalities with respect to electricity and transportation time are
also included in the final model. Internet related infrastructure at the regional level does
not show any geographical externality, which is expected since the world wide web
allows to overcome location barriers. Externality of communications appears here as one
of the top variables. Intercity extra calling costs make the case for externality for phones.
Other Externalities – If we look at total per capita income as well as labor earnings they
both show externality effects in the same fields of phone communications and
transportation. In contrast, completed years of schooling are affected by internet related
infrastructure. This may be a proxy for the effects of the digital age in schools, libraries
and so on. When we restrict this variable to school age between 7 and 15 years of age, the
main externality is yield by electricity. Programs like Light in School (Luz na Escola) and
Light for Everybody (Luz para Todos) attempt to explore this effect. Imputed rents
indicate that housing values are also affected by phone communications and
transportation costs, especially the former that occupies the top position among all
explanatory variables.
24
Proficiency, Repetition and School-Home Infrastructure
Using the microdata of the Basic Education Evaluation System (SAEB/MEC) of 2003
and 2015, we estimated the impact of infrastructure variables in school proficiency and
grade repetition. This is done combining the objective infrastructure coverage
information at students home and at school with the perceived quality of infrastructure
services in school and running regressions explaining proficiency tests and grade
repetition outcomes controlled for year (2003 and 2015), student characteristics (sex and
color), household assets infrastructure (existence of bathroom in student house and
existence of computer in student house), school characteristics (if school is private or
public and rural or urban) and school assets infrastructure (has good illumination and
well-made classrooms, has good bathrooms, water installations and electricity).
Multivariate OLS results on levels for the 5th grade in Mathematics do not allow us to
reject the hypothesis that investment in public infrastructure services is more important
for proficiency improvement than typical physical investment in school buildings, once
good electricity and water installations had a higher impact than the conservation status
of classrooms and bathrooms. Robustness tests were made with the Portuguese exams in
the 5th grade, math exams of students in the 9th grade and in the last year of high school.
Students with the same household and school characteristics had an improvement of
almost 35 points in 2015 compared with 2003, which represents a progress of the quality
of education. We observed the same pattern for similar students that differed only in terms
of infrastructure coverage, whether at home or school, as the graph below shows. Those
with access to good installations of electricity and water in school had a math proficiency,
in average, 7 and 6 points higher, respectively. It is interesting to notice that classrooms
walls in good status, our proxy for well-made classrooms, showed little importance for
the outcome, suggesting at face value that investment in public infrastructure services that
is connected outside schools was more important for proficiency improvement than
typical private investment in buildings. However, the quality of bathrooms seemed
important, once students with access to good bathrooms had proficiency 9 points higher.
The difference-in-difference method provides a dynamic analysis of the infrastructure
contribution, once it compares the difference in proficiency between students with access
to an infrastructure asset in 2015 and 2003 with the difference between the group of
25
students marginalized in terms of these assets in both years. Controlling for home and
school attributes, students with access to good electricity and water installations in school,
compared with those without that, had an average proficiency improvement of 27 and 11
points, respectively, between 2003 and 2015. In the other hand, at the same period,
proficiency of students with access to good bathrooms and classrooms in school had no
statistical difference than of students enrolled in more precarious schools. Therefore, the
diff-in-diff test corroborates the main role of public infrastructure in the recent upward
movement of school proficiency in the 5th grade.
Source: FGV Social/CPS using SAEB/IBGE microdata
#Controlled for household assets infrastructure, student general characteristics, school assets infrastructure
and year of the survey. All coefficients significant at 99%.
We also applied a process of variable selection using a stepwise statistical procedure that
ranks explanatory power of all the variables pre-selected to the model. The champion and
runner-up variables were “computer at home” and “color” of the student. “Bathroom at
home” and “local of the school” (urban or rural) were in the third and fourth positions,
respectively. Water and electricity installations were in eighth and ninth positions.
Grade Repetition and Infrastructure – To make a parallel of the present infrastructure
analysis with changes of the so-called IDEB (Basic Education Development Index), we
use the question of SAEB on grade repetition to proxy flow variables in IDEB. IDEB is
-15.0
-29.4
-3.3
9.16.4 7.2
34.6
-15.2
-31.7
-2.1
7.95.3
10.4
30.6
-35.0
-25.0
-15.0
-5.0
5.0
15.0
25.0
35.0
No ComputerHome
No BathroomHome
Badly IlluminatedSchool
Bathroom School Water School Electricity School Proficiency_Diff(2015-2003)
Proficiency Impact for Private and Public Infrastructure Assets Controlled# Multivariate Tests for the 5th grade
MATH PORT
26
a synthetic indicator of education quality based on the academic passing rate and the
results of proficiency exams (as SAEB and Prova Brasil) for each municipality and school
in the country. As we have seen, among many different social outcomes, IDEB across the
Brazilian municipalities converged at a higher speed in the last decade in the presence of
infrastructure variables, meaning the municipalities with lower initial educational
performance grew faster than the higher ones and this speed was influenced by
infrastructure. In this section, we are attempting to mimic the flow of students captured
in IDEB using the SAEB data. The main question is: Do infrastructure variables affect
grade repetition? To answer this question we generated logistic regressions using a
dummy for students that have repeated at least once. As in the previous section, our model
controls for year (2003 and 2015), student characteristics (sex and color), household
assets infrastructure (existence of bathroom in student house and existence of computer
in student house), school characteristics (if school is private or public and rural or urban)
and school assets infrastructure (has good illumination and well-made classrooms, has
good bathrooms, water installations and electricity).
Results showed statistical significant coefficients for household assets, with 12% and
37% more chances for repetition for students without computer and bathroom at home,
respectively. However, the quality of classrooms physical structure and illumination
apparently did not affect grade repetition. The only school private infrastructure with
positive impact was the quality of bathrooms, with 24% less chances for repetition for
students with good bathrooms in their schools. While water installations did not improve
school flow (with more chances of repetition for all coefficients), students in schools with
good electricity installations had 9% less chances of repeating their grade. The time
variation, measured by the dummy for 2015, suggested a marked advancement in the
education efficiency in this grade, with 95% less chances of repetition for students in
2015 in comparison with peers with the same scholar and home characteristics in 2003.
27
Conclusion
We provided an empirical analysis on the access to public services infrastructure in order
to base prescriptions for improvement policies. The final objective of this work is to create
a basic infrastructure of knowledge to guide a new generation of infrastructure programs
in Brazil. A first contribution was to analyze in a comparative way attributes of the
various public services through household surveys, such as spatial coverage, perceived
quality, expenditures and delay of accounts. We compared the coverage of these surveys
with different databases, including information provided by service providers and even
School Census, in order to more critically analyze their evolution and create monitoring
systems. The most recent evidence on infrastructure coverage in Brazil shows that the
most widespread items in 2015 were electricity (99.7%), cell phone (93.5%), water
(83.6%), private transportation (61.1%), sewerage (56.9%) and internet (42.5%),
The household survey approach is particularly useful here because it allows to study side
by side causes and social consequences of infrastructure including: Income Causality –
How much access to public infrastructure is related to exogenous increase of income;
Conditional Convergence of Infrastructure Coverage across the 27 Brazilian units of
the federation; Social Convergence analysis with and without infrastructure variables
across 5500 Brazilian municipalities – Growth regression applied to per capita GDP, per
capita household income, the Human Development Index, its 3 components plus a series
of related variables such as poverty and inequality, life expectancy, child mortality, school
attendance for various age brackets and the Basic Education Development Index (IDEB);
Distributive Impacts – Quantile regressions based platform of infrastructure impacts
along the distribution of different social outcomes of per capita household income, years
of schooling, imputed rents, the opportunity cost of commuting time and for the sake of
concision, a broader social measure that includes total reported income plus imputed rent
minus commuting costs; Infrastructure Externalities – A stepwise variable selection
procedure to determine which socio-economic and infrastructure related variables
included externality effects from infrastructure are more statistically important to explain
each social outcome analyzed. School Proficiency SAEB/MEC tests were also used to
test the impact of school and home infrastructure on school performance.
28
Infrastructure Investment and Social Progress in Brazil
Full paper
1. Introduction
There are many public infrastructures and associated universalization policies impacts.
We can cite the so-called social infrastructure items such as basic education and health.
There is also urban transportation, information and communication technologies (ICTs)
and a myriad of public services regulated by state agencies offered by municipalities, or
privatized companies in sectors such as electricity, water and sewerage, among others.
This diversity of arrangements plus their interaction suggests a vast array of possibilities
on the analysis of the causes and consequences of investment in infrastructure. This paper
draws a broad empirical diagnosis on the level and on the evolution of infrastructure
coverage in Brazil and their potential social impact.
The present study focuses attention on the sectors of sewerage, water, electricity, urban
transportation and Information and Communication Technologies (ICTs). We develop
most of the analysis departing from household surveys, bringing the population
perspective into the picture and exploring possible public policy implications. We take
advantage of the many dimensions offered by the microdata sources to develop bivariate
and multivariate type of analysis of socio-economic determinants of infrastructure
coverage, which can be seen as a social outcome in itself, as well as their possible social
impacts. The latter manifestations include flows such as income and earnings generation,
time cost of transportation and stocks such as housing value and education outcomes.
In all stages of the analysis we look for causes and consequences of coverage, static and
dynamic, bivariate and multivariate, emphasizing the roles played by two dimensions: per
capita income and the geography in order to locate necessary public policies efforts in the
social strata and in the territory. In particular, we emphasize the role of Brazilian States
as a unit in the analysis. For example, in the case of the coverage level, multivariate
analysis allows comparing access of individuals with the same observable characteristics
(income, family size, education, gender, city size etc.) across different units of federation.
This allows us to map repressed demand for a future infrastructure expansion. By the
same token, we also compare the relative evolution of these type of individuals in the
same areas across time using a difference in difference approach to check if there is a in
infrastructure coverage convergence process going on. The income dimension enters both
as a possible determinant as well as a consequence of infrastructure coverage. We test
29
how much it affects the income convergence across Brazilian spatial units and how much
“exogenous” household income increases affect infrastructure outcomes.
Script – This paper undertakes a broad empirical description on causes and consequences
infrastructure coverage in Brazil. It extends our previous work on infrastructure sectors.
The script of this paper is the following: section 2 introduces the conceptual framework,
data sources and estimation procedures used in the paper. It also illustrates with the most
recent data the coverage level observed in the country. Section 3 describes the evolution
of infrastructure coverage in Brazil. In most of the analysis, we look at a comparative
perspective across infrastructure sectors. We explore maps at State and Municipal levels
keeping the scales constant across time. Section 4 also explores bivariate dimensions of
coverage such as income, age, city size and region. Section 5 attempts to isolate each
socio-demographic dimension in coverage using multivariate estimation methods,
typically arising from logistic regressions. We analyze the spatial convergence of
infrastructure keeping determinants constant across time. Finally, in search of causality
direction, it explores the expansion of Bolsa Família as an experiment of the impacts of
pure income effects associated with the expansion of anti-poverty policies on public
infrastructure service coverage. This is a key policy related point of the article. The
exercise shows that income increases are not always accompanied by more infrastructure.
Section 6 starts to analyzing possible social consequences of increasing infrastructure
coverage in the budget constraint taking into account services costs and payments delays.
This section also incorporates subjective perceptions on the quality and importance
attributed comparatively to different infrastructure sectors. Section 7 test using municipal
level data, the role of infrastructure in the spatial convergence of income and other social
dimensions such as poverty, human development and its components. Section 8 explores
the same issue in a more general setting using microdata and quantile regressions. We
estimate the potential impacts of different infrastructure sectors along the distribution of
various social indicators such as per capita total income, labor earnings, education,
imputed rents, commuting costs and a constructed broader social welfare measure. Section
9 applies a variable selection procedure to rank infrastructure variables in terms of
potential social impact, such as on poverty and on those discussed in section 8. It includes
as well externality effects of infrastructure at the community level on individual social
outcomes. This market failure may justify as well as signal the necessity of certain policy
interventions. The last section summarizes our main conclusions.
30
2. Empirical Diagnosis on Infrastructure Coverage3
This section introduces the conceptual framework, data sources, estimation procedures
used in the paper and presents recent data on the level of infrastructure coverage.
Motivation – Policy-makers and researchers of the problems of emerging countries,
particularly in the case of China and South Africa, have recurrently used the term
"Brazilianization" as representative of the disordered growth of large cities. Over the last
century, Brazil has become an essentially urban country, with 85% of the population
living in cities. According to the Census of 1940, 31.2% of our population lived in cities,
according to the last PNAD, collected in 2015, almost the same proportion of people,
31.5%, live in metropoles and 54.9% lives in other urban areas. Throughout this process
of urbanization, we have learned the costs of diseconomies associated with this Brazilian
population agglomeration, such as chaotic traffic, informality in access to infrastructure,
the impact of these bottlenecks on productivity growth, education and the unhealthiness
of our daily living conditions. On the contrary, we should offer more and better public
services by exploiting the economies of scale, scope and network, for having a large part
of the population in these large cities. That is, large cities should not be synonymous with
precariousness, visible in the favelas and peripheries that stand out today as images of the
country alongside the recent fall of the economy.
The urban disorder of the Brazilian case surprises more than the Indian one, because we
have more income and a larger State. However, these are not enough conditions to avoid
the chaos of cities through more investments in infrastructure, even if accompanied by
reduced income poverty and inequality. The incentives framework for consumers and
service providers is necessary in order to flourish social infrastructure and logistics. The
clearest example of Brazil's waste of opportunity is basic sanitation. However, even the
largest Brazilian cities - given the location of the population - do not enjoy this basic item.
We live in the 21st century as if we were in a 19th century European city. The exception
is the universalization of electricity in cities, where the problem is concentrated in non-
technical losses. Urban transportation measure in terms of commuting time got worst as
a collateral effect of the previous boom in a context of absence of the supply of public
3 This work extends a series of research carried out by FGV Social for the World Bank and also benefits from previous research performed in sewerage and water for the NGO Trata Brasil, on ICTS for Telefonica Company and on Urban Transportation for the State of Rio de Janeiro. (see WWW.fv.br/cps/tratabrasil5).
31
means of transportation. Internet and especially cell phones expansion opens new
possibilities for making our cities smarter. But this is the story we are going to explore in
detail. First, we need to get acquainted with the conceptual approach pursued here, the
sources of microdata used and the techniques applied throughout the paper.
Conceptual Framework - The potential social impacts channels of infrastructure changes
will be assessed here under three headings. First, the direct impact on well-being modelled
by the individual utility function. Although one may assume different degrees of
substitution or complementarity between different infrastructure items and other variables
such as income, we will solely interpret directly the data on coverage rate - as a direct
social impact in itself. This is the main line of inquiry pursued here. Additionally,
subjective question on the satisfaction level obtained will add a quality measure of
infrastructure coverage. We also will interpret directly subjective questions on the
importance assumed by specific elements as another measure of the relative importance
of direct well-being effect across different infrastructure sectors.
Second, the direct impact on the monetary budget constraint depending on the way
infrastructure services are financed. Household expenditure surveys offer evidence of this
channel operation by capturing the size of infrastructure pay bills. Also questions on the
delay of this payment bills will add evidence on this current budget constraint effect and
help the design of public policies. This channel evidence will be somewhat limited by the
scarcity of more recent data.
A third channel explored here is the influence exerted by infrastructure on individuals
income and assets generation process, modeled by movements along the production
function, shifts in the production function and on the way individuals connect to inputs
and outputs markets - for example, in the case of transportation and ICTs. We will not
attempt to disentangle empirically the operation of these productive/market channels
labelling them broadly as income generation. At an intermediary level of aggregation, we
perform standard growth regressions to test how much infrastructure adds to socio-
economic convergence between Brazilian States and between Municipalities, meaning not
only income convergence but also Human Development Index components convergence
across these units. Then departing from individual data, we will recur to the estimation of
mincerian log-linear income equations for the mean and quantiles to estimate the
correlations across the whole distribution. We incorporate in these income regressions the
32
interaction with human capital and social economic characteristics. Hedonic rent
equations add evidence of the infrastructure effects on assets value since housing turns out
to be the most important physical asset of the family in Brazil and elsewhere. Similarly,
we also perform exercises for the time opportunity cost of transportation and for broader
societal well-being indicators that includes all the above. Equations on the impact of
infrastructure on years of schooling and proficiency will complement the analysis.
Data sources - The main Brazilian National Household Survey, PNAD/IBGE, will be the
key source of data used along the paper, including the very last edition of PNAD recently
made available by IBGE. PNAD was discontinued so in a sense this paper consolidates
the historical series of the survey. We use the survey data from 2004 and 2015,
complementing previous work done for the World Bank. This is also the period when
PNAD has a national coverage, including the rural area of the North region of Brazil. The
two last versions of the Demographic Census provide a finer geographical definition of
coverage rates and their social consequences. Providing more degrees of freedom to
estimate the impact of infrastructure on income convergence across Brazilian
municipalities. We will use other sources of microdata such as Consumer Expenditures
Survey (POF/IBGE) to capture direct household budget impacts of infrastructure
provision as well as qualitative assessment of infrastructure. This subjective approach will
also be pursued using other sources to capture priorities among infrastructure sectors and
demand related motivations. We will apply a broad set of microeconometric techniques to
these microdata sets including logistic regressions, mincerian income equations, quantile
regressions, difference in difference estimators and stepwise variable selection applied on
top of these empirical models. The data sources and microeconometric techniques used
are all described in the appendix.
Coverage Level - We analyze the most recent evidence on the level of infrastructure
coverage in Brazil as a whole. We focus initially in their respective coverage rates from a
household perspective using simple proxies that can be used during the 2004 to 2015
period. The most widespread items in 2015 were electricity (99.72%), cell phone
(93.46%), water (83.58%), private transportation (61.09%), sewerage (56.89%) and
Internet (42.47%), We present below the rates of coverage opened by the 27 Brazilian
Federation Units using the same scale that confirms that sewerage presents not only a low
but also highly variable coverage across Brazilian States.
33
Same escale Link – 2015 (%) Has a sewerage system Link - 2015 (%) Water (%)
Has cell phone - 2015 (%)
Has internet access - 2015 (%)
Electricity - 2015 (%) Daily one-way journey time To the workplace (Hours)*
Source: FGV Social/CPS from the PNAD/IBGE microdata Same scale except*
34
3. Evolution of Infrastructure Coverage
Next, we analyze the evolution of total coverage of the population and by the different
segments of society trying to identify their closest determinants. The period of analysis
cover from 2004 to 2015, in which PNAD offer a representative sample of the country as
a whole including the rural areas of the North region. The graphs below display the main
changes in the coverage rate or related statistics of this group of six infrastructure items.
In general, we observe a rise in the infrastructure coverage
Infrastructure Coverage Evolution % – Public Services
Infrastructure Coverage Evolution % – ICTs
45,06 45,3 45,4448,05 49,76 49,76 51,96 54,27 55,35 55,36 56,89
77,4 77,68 78,94 79,72 80,69 81,46 81,98 83,13 82,57 83,42 83,58
96,27 96,57 97,17 97,89 98,33 98,73 99,24 99,47 99,54 99,67 99,72
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Sewerage Network Water Network Electricity
47,81
60,3964,85
69,11
77,7380,65
88,62 90,54 91,88 93,41 93,46
11,47 13,1616,27
19,7623,92
27,43
37,1941,55 43,84 44,08 42,47
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
CellPhone Internet
35
Source: FGV Social/CPS from the PNAD/IBGE microdata
Access to infrastructure services has increased significantly over the past decade.
This is mainly due to lagged effects of the privatization programs of the 1990s (especially
in telecommunications), the adoption of public programs aimed at expanding coverage in
remote areas (especially in electricity due to the “Luz Para Todos” program) and the
demand effect from the combination of faster household income growth and falling
inequality that lasted until 2014. Using household level data on coverage of infrastructure
services, the service that had the highest increase in access between 2004 and 2015 was
ICT. The past 10 years has seen an explosion in the use of mobile telephones. In 2004,
around 85 million people had mobile phones at home, and in 2015 the number increased
to 186 million – an increase of 101 million users. During the same period, home internet
coverage was extended to an additional 64 million Brazilians. Despite its rapid growth,
internet service is the infrastructure service that presents the lowest level of access (42.5
percent) when compared to other services. On the other extreme is electricity, with an
access level of 99.7 percent. Access to potable water has an intermediate rate of 83.6
percent, but significantly more than sewage services, at 56.9 percent.
1,00
1,05
1,10
1,15
1,20
1,25
1,30
1,35
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Mean Commuting Time (Hours)
Commuting Time
36
Brazil: share of population with access to infrastructure services (%)
Source: FGV Social/CPS from the PNAD/IBGE microdata
11,5
45,1
77,4
47,8
96,3
42,5
56,9
83,693,5 99,7
0
20
40
60
80
100
internet sewerage water cellphone electricity2004 2015
37
4. Bivariate Analysis of Infrastructure Coverage Evolution
While there has been some convergence over the past decade, significant regional
differences remain across the country in terms of access to infrastructure services,
particularly in water, sewerage and internet services. The states with the highest rate of
access are in the Southeast and South: São Paulo, Santa Catarina. Households in the
Federal District also enjoy high levels of access to infrastructure services. There is more
variability among the lower levels of the rankings, but states from the North and Northeast
regions tend to be at this end of the spectrum. In terms of internet services only 15% of
the population in Maranhão and Pará have home access compared to 67 percent in the
Federal District – a more than 50 percentage point difference between extremes. In the
water sector, access also varies considerably across the different states. In São Paulo,
access to the water network is around 96 percent, while in Rondonia, access does not
reach half of this proportion (Figure 5). With respect to sewerage that inherits some of
the water attributes, 91 percent of São Paulo has access and only 8 percent has access in
Rondonia. In contrast to the other infrastructure services, electricity coverage displays a
more homogeneous spatial distribution with at least 99.99 percent of the populations of
São Paulo, Distrito Federal and Rio de Janeiro having access and on the other extreme
around 95.5 percent of households in Acre have access.
Coverage of infrastructure services in rural areas has expanded but the sharp divide
between rural and urban coverage within the country persists. Only in sanitation has
rural coverage not changed much. However, access gaps between rural and urban areas
remain high. While rural areas represent around 14 percent of the Brazilian population in
2015, only 4 percent of this population has access to sewerage services with only a third
having access to the water system. In urban areas, where most of the population lives, the
rate of access to the water system is about 90 percent, while access to sewerage services
is about 80 percent. The pattern of low rates of access in rural areas and high rates of
access in urban areas is evident in all infrastructure services with the exception of
electricity, where access rates have converged.
Infrastructure access reflects and reinforces Brazil’s high income inequality. Access
rates among the poor have been improving in the last decade but coverage remains much
higher among wealthier groups.. Sewerage, water and internet tend to be the most
unequally distributed services across income groups. In 2015, less than half of the poorest
38
segment of the population had access to sanitation facilities, compared with 80 percent of
the richest.
Income Group – % Infrastructure Coverage
Electricity Sewerage Network
Water Network Home Internet Access
Cell phone Car
Source: FGV Social/CPS from PNAD/IBGE microdata
88
90
92
94
96
98
100
102
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Electricity
Bottom 40%" 40% to 90% Top 10%
0
10
20
30
40
50
60
70
80
90
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Sewerage
Bottom 40% 40% to 90% Top 10%
0
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Water
Bottom 40% 40% to 90% Top 10%
0
10
20
30
40
50
60
70
80
90
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Home Internet Access
Bottom 40% 40% to 90% Top 10%
0
20
40
60
80
100
120
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Cellphone
Bottom 40% 40% to 90% Top 10%
0
10
20
30
40
50
60
70
80
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Car
Bottom 40% 40% to 90% Top 10%
0100
40% less 40% to 90% 10% plus
39
Age – % Infrastructure Coverage
Electricity Sewerage Network
Water Network Home Internet Access
Cell phone Car
Source: FGV Social/CPS from PNAD/IBGE microdata
92
93
94
95
96
97
98
99
100
101
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
0
10
20
30
40
50
60
70
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
65
70
75
80
85
90
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
0
10
20
30
40
50
60
70
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
0
20
40
60
80
100
120
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
0
5
10
15
20
25
30
35
40
45
50
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
050
0 to 4 5 to 9 10 to 14 30 to 35
36 to 39 55 to 59 60 years or +
40
Type of Area – % Infrastructure Coverage
Electricity Sewerage Network
Water Network Home Internet Access
Cell phone Car
Source: FGV Social/CPS from PNAD/IBGE microdata
0
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Electricity
Metro cities Urban non metro Rural
0
10
20
30
40
50
60
70
80
90
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Sewerage
Metro cities Urban non metro Rural
0
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Water
Metro cities Urban non metro Rural
0
10
20
30
40
50
60
70
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Home Internet Access
Metro cities Urban non metro Rural
0
20
40
60
80
100
120
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Cellphone
Metro cities Urban non metro Rural
0
5
10
15
20
25
30
35
40
45
50
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Car
Metro cities Urban non metro Rural
Metro cities Urban non metro Rural
41
Macro-Regions – % Infrastructure Coverage
Electricity Sewerage Network
Water Network Home Internet Access
Cell phone Car
Source: FGV Social/CPS from PNAD/IBGE microdata
0
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Electricity
Nordeste Sudeste Sul
0
10
20
30
40
50
60
70
80
90
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Sewerage
Nordeste Sudeste Sul
0
10
20
30
40
50
60
70
80
90
100
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Water
Nordeste Sudeste Sul
0
10
20
30
40
50
60
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Internet Access
Nordeste Sudeste Sul
0
20
40
60
80
100
120
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Cellphone
Nordeste Sudeste Sul
0
10
20
30
40
50
60
2004 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015
Car
Nordeste Sudeste Sul
Nordeste Sudeste Sul
42
Bivariate Analysis - public services coverage is crossed by various dimensions, such as:
income, age, type of area and macro region. As a general remark graphs on the left such
as electricity, water and cell phone tend to present a faster rate of convergence than those
on the right side such as sewerage, internet and cars.
i) income per capita - given the emphasis on combating poverty and inequality, plus the
possibility of subsidies on income brackets, we choose to divide the sample in three
groups: the bottom 40%, which is aligned with 11th target of the United Nations
Sustainable Development Goals, the top 10%, given their explanatory power in Brazilian
income distribution, and the intermediary group between these two extremes, which can
be seen as a sort of relative middle class in a statistical sense. The income dimension tends
to reproduce the sharper rate of convergence for electricity, water and cell phone,
mentioned above.
ii) age - providing a long-term view of how different age groups benefited or not from
this coverage, also emphasizing the extremes of the distribution. It is impressive the
division by age of coverage in traditional public services and cars, where children have a
much smaller access. While in ICTs the age division is much less pronounced. For that
matter, the elderly tend to have lower ICT access in spite of their higher income levels.
iii) type of area – including the division between metro cities and other urban areas, which
may offer economies, or diseconomies, of scale. The rural area has only a sharp
convergence movement in the case of electricity and to lesser extent, in cell phone
coverage.
iv) macro-regions – In particular the contrast between the two most populated regions of
the country: the rich Southeast and the poor Northeast. The south tend to follow the
Southeastern levels. One regional feature pointed in previous studies is the smaller access
to sewerage network in the rich South part of Brazil, where in spite of some recent catch
up movement, its rates of exclusion are almost at Northeastern levels.
We devote now our efforts to map the evolution of the geographical distribution of
infrastructure items across Brazilian States between 2004 and 2015 using the same scale
across time.
43
Electricity – 2004 / 2015 (%) (http://cps.fgv.br/tem_Has Access to Electricity)
Source: FGV Social/CPS from the PNAD/IBGE microdata
44
Water - 2004 / 2015 (%) (http://cps.fgv.br/tem_agua_2004_2015)
Source: FGV Social/CPS from the PNAD/IBGE microdata
45
Has a sewarage system - 2004 / 2015 (%) (http://cps.fgv.br/tem_esgoto)
Source: FGV Social/CPS from the PNAD/IBGE microdata
46
Has cell phone – 2004 / 2015 (%) (http://cps.fgv.br/tem_celular_2004_2015)
Source: FGV Social/CPS from the PNAD/IBGE microdata
47
Has Home internet access – 2004 / 2015 (%) (http://cps.fgv.br/computador_com_internet_2004_2015)
Source: FGV Social/CPS from the PNAD/IBGE microdata
48
Daily one-way journey time to the workplace – 2004 / 2015 (Hours) (http://cps.fgv.br/tempo_de_transp_2004_2015)
Source: FGV Social/CPS from the PNAD/IBGE microdata
49
Individuals in Households with Car (%) (http://cps.fgv.br/carro_carro_mais_moto)
2008 ________
2014
Source: FGV Social/CPS from the PNAD/IBGE microdata
2010
50
Individuals in Households with Motorbike (%) (http://cps.fgv.br/porcentagem_domicilios_com_moto_escala_conjunta)
2008 _______
2014
Source: FGV Social/CPS from the PNAD/IBGE microdata
51
5. Determinants of Infrastructure Coverage
Multivariate exercises – We ran now logistic regressions to isolate the determinants of
infrastructure coverage in the period of analysis. Besides gender, race and spatial variables
we use second degree polynomials for per capita income, family size, education and age.
These quadratic terms turned out significant in most of the regressions.
Year effects - The regression analysis allows us to measure the growth rate of odds ratio
between 2004 and 2015 keeping socio-demographic structure constant. The highest
temporal change was observed in electricity, internet and cell phone. The lowest
expansion was found in water and sewerage.
Family size – This variable present in general a positive but at diminishing rates effect.
This point is noteworthy since as product of the demographic transition household size
has been decreasing. For example, between 2004 and 2015 the mean number of members
per family was reduced from 4.38 to 3.74, a 14.7% total fall. Population size grows now
in Brazil at a 0.8% per year while the household size falls 1.43% per year, creating an
additional pressure on the infrastructure supply. This means that the infrastructure supply
has to increase not only because of the existing infrastructure deficit and population
growth but also because the number of dwellings also increased as a response to the
household size reduction effect, requiring new infrastructure connections.
The per capita income effect is positive and diminishing in general. But causation is
not warranted in any of these partial correlations. Given its central economic meaning it
is worth analyzing a quasi-experiment presented further below.
States Evolution – Many of the spatial differences of infrastructure coverage may be
attributed to differences in income, education, family size, city size, states and so on. In
order to net out these influences, we use multivariate regressions of coverage described
above. We focus our analysis on the later spatial variable. The maps presented in each
page present the geographical dispersion of coverage across Brazilian states. São Paulo is
always portrait white as the basis (i.e. the omitted variable). The red means that is lower
than São Paulo, while blue gives the excess with respect to São Paulo. As a general rule,
52
all other States appear in different tones of red except for some statistical draws, meaning
that the State of São Paulo presents the best infrastructure in the country4.
Next we run an extension of the previous multivariate exercise also incorporating the
interaction between State Dummies and year in order to grasp the spatial dimension of
infrastructure coverage changes. In this second type of regression, we fixed São Paulo as
the omitted spatial dummy and 2004 as the omitted temporal category. In this way the
results are directly interpreted as the conditional difference in difference of each state in
2015 with respect to São Paulo in 2004. Or how much the infrastructure coverage changed
in relative terms. In most cases the color of the map turns into blue which means that the
differential between different states and São Paulo tended to fall. This shows a clear
convergence trend of infrastructure between Brazilian states even if we net out the effects
of income, education and other variables during this period. To be sure, comparisons
among states show that an individual from São Paulo has the highest chance of having
access to almost all infrastructure services than a similar individual in any other state of
the Brazilian Federation. When we move to the comparison of movements of coverage
rates, in most cases the color of the map turns into blue. This means that the differential
between different states and São Paulo tended to fall. This suggests a clear convergence
trend of infrastructure between Brazilian States even if we net out the effects of income,
education and other variables during this period.
Details: Taking São Paulo as the basis, the convergence movement is true for basic public
services such as Electricity and Water. Electricity convergence exceptions was found in
the States of Amazonas and Roraima and in the case of Water Network the State of Amapá
was the sole exception. The location of these States in the more remote areas in the
Brazilian Amazon are probable driving forces behind these exceptions . Sewarage,
internet and telephony presents 5 exceptions among 27 states to the general rule of
convergence with respect to São Paulo. São Paulo comes in second place and Tocantins
in last, with 95 percent less of a chance than a similar person living in São Paulo. Access
to both mobile and fixed telephone services are better distributed among the different
states. The states with higher probability of access to these services are Distrito Federal
and Rio Grande do Sul, and the states with the lowest probability are Pará and Ceará.
4 For example, Rio de Janeiro in the case of electricity.
53
Electricity - States Odds Ratio
http://cps.fgv.br/razao_Has Access to Electricity
Electricity – States Odds Ratio with Time Interaction
http://cps.fgv.br/razao_Has Access to Electricity_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
54
Water (General Network) - States Odds Ratio States Odds Ratio
http://cps.fgv.br/razao_agua
Water (General Network) - States Odds Ratio with Time Interaction http://cps.fgv.br/razao_agua_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
55
Has sewerage - States Odds Ratio
http://cps.fgv.br/esgoto_razao
Has sewerage - States Odds Ratio with Time Interaction http://cps.fgv.br/esgoto_razao_com_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
56
Has a computer with internet at home - States Odds Ratio
http://cps.fgv.br/razao_comp_com_net
Has a computer with internet at home - States Odds Ratio with Time Interaction http://cps.fgv.br/razao_comp_com_net_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
57
Has a landline or mobile phone at home - States Odds Ratio
http://cps.fgv.br/razao_fixo_celular
Has a landline or mobile phone at home - States Odds Ratio with Time Interaction http://cps.fgv.br/razao_fixo_celular_interacao
Source: FGV Social/CPS from the PNAD/IBGE microdata
58
Ln equation of course time evaluated at hourly-wage (main job) - States Odds Ratio http://cps.fgv.br/estimativa_equacao_de_ln_de_tempo_de_percurso_ao_trab
Ln equation of course time evaluated at hourly-wage (main job) - States Odds Ratio with Time Interaction
http://cps.fgv.br/interacao_equacao_de_ln_de_tempo_de_percurso_ao_trab
Source: FGV Social/CPS from the PNAD/IBGE microdata
59
Causality and the Bolsa Família Experiment – Next, we used the marked expansion of
Bolsa Família between 2004 and 2006, when it almost doubled the number of
beneficiaries, to test the effect of “exogenous” income changes on access to public
services. For this, we use the 2004 and 2006 PNAD supplements on social programs. The
effect is captured by a difference-in-difference estimator generated from the interaction
of the dummy variable year (before and after expansion) with the dummy variable for the
program eligibility criterion (per capita household income less than R$100 monthly in
real terms, excluding income earned by social programs). The regression is also
controlled by age, race, migration, and other variables, such as a dummy for a slum
dweller, demographic density, and federation unit.
We present the results of the multivariate logistic regression models of access to different
services to try to capture the effects of the income expansion, using as an instrument the
population eligible to Bolsa Família, controlling for the same characteristics mentioned
above5. That is, we analyze how much the increase of the access to public services is
related to the increase of income of this population through the expansion of 67% of the
program coverage between 2004 and 2006. The following results focus on the variables
used in the interaction, isolated and combined. These variables show that, in the
controlled analysis, electricity, garbage, cell phones and internet grew in the period: The
access chance is 2 times greater in the second year. Sanitation, water and landline
telephone services have a relative drop (odds ratio of 2006 in relation to 2004 of 0.97,
0.96 and 0.79, respectively) when we control for the attributes of the person. In the case
of transportation time we use a log-linear regression using that same controls. The results
shows an increase of 1.1% between 2004 and 2006. Next, we compare the access of the
eligible population to Bolsa Família versus the others with all similar characteristics,
including income as a continuous variable: the chances of access to all these services and
assets, except for the general water network, are lower for the low income group. In the
case of access to sanitation, the odds ratio of the low income in relation to the others is
0.71. The transportation time was 1.2% higher for the low income group reflecting the
impact of lack of resources on the outcome. Finally, we used the dummy variables above
(eligible*year) to measure whether the income gain of the low-income population
increased more than the others. The results are a relative improvement for all items
5 Neri and Andrade (2011) presents a description of the logistic regression technique used here and the estimated complete models.
60
(except sanitation). In the case of cellphone access and fixed telephone the chances are
13% and 11% higher, while in access to public services, such as garbage collection,
electricity and general water network, the chances are 13%, 11% and 8% higher,
respectively. The improvement of transportation is captured by a -1.3% fall in commuting
time at individual level6. The same goes for assets such as computer connected to the
internet and bathroom at home. However, for sewerage connected to network there was
no statistically significant improvement in relation to the other group. The higher income
did not impact access to the sewerage network of the population eligible to the program.
This lack of sensibility may be due to the predominance of externalities in the supply of
sewerage where individual or private returns to sewerage connection benefits mostly
others.
6 We run a similar exercise using transportation time evaluated at the wage rate of the commuter and there was no statistically significant change between these groups.
61
6. Perceptions, Priorities and Empirical Comparisons
We also incorporated a more detailed geographical analysis. In household surveys, we
included Census, which provides a longer, more spatially detailed view; the National
Household Sample Survey (PNAD), which provides the temporal details and updates this
evolution; and the Family Budget Survey (POF) that allows the measurement of impacts
in the household budget and the perceived quality of services.
Empirical Comparisons - One advantage of using household surveys such as PNAD and
POF is to analyze people's views. We can also use data from service providers through
the National Sanitation Information System (SNIS) and data on water and sanitation
reported by companies to the Ministry of Cities. Barely comparing, while the latter
analyze people more informed and interested in the subject, the latter analyze more
uninformed people, I admit, but also more disinterested in appearing good or bad in
statistics. The two pieces of information are complementary. We propose here a
conciliation: to use the 2008 School Census information on 197 thousand Brazilian
schools. School principals are more informed than the average citizen who responds to
household surveys, but also more disinterested to appear good than the manager of a
service provider. There is reasonable consistency between rates of coverage of public
services in schools and those perceived in households, at least within the capitals of the
federation units.
Comparison of Coverage in Schools - The results presented now reflect what we
observed in Brazilian schools, in which the lack of sanitation is more intense than of other
public services. While the proportion of schools with sanitation in 2008 was only 39.58%,
the other services coverage are much higher: water supply (62.64%), electricity (88.24%)
and Garbage collection (62.93%). It should also be noted that sanitation in schools is
lower than for the households. One advantage of the School Census is to allow the yearly
analysis of various infrastructure items at the municipal level7.
7 SAEB microdata also from MEC allow us to monitor every two years at more aggregate State level, but
also including home coverage data and questions related to the school infrastructure quality perception.
This data will be analyzed in section 10.
62
Perceived Quality - The IBGE Household Budget Survey allows us to explore the
perceived quality of access to services. That is, we leave the dichotomy between having
and not having access to sanitation or water and enter into the subjective scope. In general,
the quality of services associated with water enjoys lower perceived quality than that of
public services such as electricity and garbage collection. Regarding access to water,
82.5% of the Brazilian population evaluates access as "good" and the rest consider it
"bad", while only 71% of those who have access to sanitation consider it "good". For
electricity and garbage collection services, the percentages for "good" are 92.45% and
87.65%, respectively. It is worth remembering that we are only evaluating quality here,
not the percentage of access.
Perceived Quality in Metropolis – Infrastructure supply is heavily influenced by
economies of scale involved in the construction of networks. An analysis at the main
Brazilian cities level should yield a more relevant context of comparison among various
public services. We observed that the level of general sewage network coverage in the
metropolis (67.5%) was much lower than other public services, such as water (92.3%),
garbage (86.8%) and electricity (98.2%). Note again that general sewage network
coverage is a necessary condition for the provision of sanitation, which in turn is a
sufficient condition for the collection benefits to materialize in their integrity. The same
is true for the perceived quality of public services in schools. In general, the quality of
services associated with water enjoys lower perceived quality than that of public services
such as electricity and garbage collection. Regarding access to water, 81% of the
population living in a metropolis evaluates access as "good" and the rest consider it "bad",
while only 69.5% of those who have access to sanitation consider it "good". For electricity
and garbage collection services, the percentages for "good" are 92.3% and 87.8%,
respectively.
The answer to the emphasis given to basic sanitation bad indicators is due not only the
lower level of coverage and perceived quality of sewage, or the lower rate of relative
growth of this service over time. Is represents an opportunity we have to begin to change
now, in a more quickly way, the sanitation framework, which is a function of the advent
63
of the new regulatory framework, with more resources available and greater awareness
of the population and the political class in the cause of sewage8.
Priorities – Besides the subjective quality attributed to each infrastructure service, one
may also investigate the weights given to them by the population itself. An analysis of
the priorities of the Brazilian population is made in terms of public policy vis-à-vis the
global population through the questionnaire in My World, from the UN to support the
definition of the new Sustainable Development Goals. These questions were incorporated
in an national wide representative household survey implemented by Ipea in 2013 with a
sample size of 3,8 thousand individuals across 215 Brazilian cities. Out of 16 items,
infrastructure variables stay in the following positions: Transportation (7th); Water and
Sanitation (9th); Electricity (13th) and ICTs (16th). According to the global wide sample
infrastructure priorities were: Water and Sanitation (5th); Transportation (12th); ICTs
(16th) and Electricity (15th).
Expenditures and Delays in Accounts - Household per capita expenditure with water
and sewage bills for each Brazilian is R$4.48 per month at December 2008 prices (65.5%
of the population has expenses with these services, which represent 0.79% of the labor
earnings). Among those who actually have these expenses the expenditure is R$ 6.83 per
capita per month. The values of these accounts are slightly higher in the total capital
population than in the peripheries: R$ 5.54 against R$ 5.1 in per capita terms per month,
respectively. This occurs even with a lower proportion of the population with this type of
expenditure in the former versus the latter, 66.5% compared to 70.3%, respectively.
The POF also allows analyzing delay of light, gas, water and sewage taken together. It
was found that, of the sample among those with water and sewage bills, 45.65% delayed
household bills in the last 12 months. The delay was reported higher in the capitals than
in the suburbs, 51.5% and 48% 7%, respectively. These problems of delay can inhibit and
even prevent the provision of the service by the operators. The other economic issue here
are the so-called technical loses involved in the provision of public services as a result of
informality in the access of public services, in particular electricity.
8 Studies show that for every R$1 applied on sanitation there is an economy between R$1.5 and R$4 in the health system expenditures.
64
7. Social Convergence and the Role of Infrastructure
In this section, we followed the standard economic growth literature and tested the role
of infrastructure variables in terms of reducing inequality between income and other
variables across Brazilian Units of the Federation. We basically implemented a standard
convergence analysis running regressions of income growth against the natural logarithm
of initial per capita household income comparing the results with and without
infrastructure variables. Preliminarily, we use the same infrastructure variables described
above from PNAD. using the 27 Brazilian states as units of analysis from 2004 and 2015.
The results were not very satisfactory given the few degrees of freedom involved in this
type estimation.
The next step was to try to overcome this lack of degrees of freedom scarcity using data
at municipal level with more than 5500 observations for each year. We decided to extend
the analysis beyond income using a myriad of social endogenous variables coming from
various sources: municipal accounts, Atlas of Human Development (Ipea, UNDP and FJP
(2013), proficiency data from INEP plus a series of infrastructure constructed using the
Demographic Census microdata. The set of variables tested includes per capita GDP, per
capita household income, the Human Development Index, its 3 components plus a series
of related variables such as poverty and inequality, life expectancy, child mortality, school
attendance for various age brackets and the Basic Education Development Index (IDEB)
which includes the results of proficiency exams.. For each of these variables we run a set
of three regressions.
i) Unconditional growth regression: endogenous variable growth against the
natural logarithm of its initial value plus a constant. This simple regression
provides useful references for the comparison with the next two regressions.
ii) Conditional growth regression: same as above (endogenous variable
growth against the natural logarithm of its initial value) plus a set of
infrastructure coverage variables which includes water, sewerage, garbage,
electricity, telephone and computer in 2000. The comparison of the lagged
variable coefficient with the first regression gives an idea of the infrastructure
variable in terms of the observed speed of convergence. While the comparison
65
of the adjusted R2 yields the marginal contribution of infrastructure variables
in terms of growth.
iii) Simple growth against infrastructure regression against the six
infrastructure variables plus a constant: the R2 provides de gross explanatory
power of the set of infrastructure variables by itself.
We present in the graph below a synthesis of the speed of convergence for all variables.
For 16 out of the 17 endogenous variables tested, the speed of convergence is higher at
face value with the set of infrastructure variables than the model without infrastructure.
Source: FGV Social / CPS from the Demographic Census IBGE microdata; Ipea,
UNDP and FJP (2013) and INEP/MEC.
The somewhat surprising exception is GDP, where the size of the lagged endogenous
variable coefficient decreases in absolute terms when we include the infrastructure
parameters. Household income seems to capture better than GDP the infrastructure
induced effects.
# This regression is for the endogenous variable in percentage against its variation in percentage points
## sample for 5010 cities between 2007 and 2015
-4 -3 -3 -2 -2 -1 -1 0
Basic Education Index (IDEB) for the 5th grade##
Basic Education Index (IDEB) for the 9th grade##
School Attendance - Children 4-6 years
School Attendance - Children 6-14 years
School Attendance - Children 6-17 years
Life Expectancy
Gini Index
Poverty (Proportion of Poor)#
Child Mortality Under 1 year
Child Mortality Under 5 years
HDI Income Component
HDI Health Component
HDI Educational Component
Human Development Index (HDI)
Per capita Household Income
Per capita GDP
Regressions for Rates of Change across 5500 Municipalities between 2000-2010LN (Endogenous Variable) Coefficient
LN (Endogenous) + Infrastructure variables LN (Endogenous) only
66
Poverty rate regression was treated in levels with the results showed below:
-1
-0.5
0
0.5
1
1.5
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
Esti
mat
ed p
er c
apit
a In
com
e V
aria
tio
n 2
00
0-2
01
0
LN (per capita Income in 2000)
Convergence in per capita Household Incomebetween Brazilian municipalities
Y estimated only with LN(pc Income) Y estimated w/ infrastructure variables
Endogenous Variable: Poverty (Proportion of Poor) Variation 2000-2010
Coefficients Stat t Coefficients Stat t Coefficients Stat t
Intercept -6.991 -39.45 11.029 14.39 -19.622 -35.55
Endogenous -0.264 -69.88 -0.333 -49.89
Water_network -0.015 -3.13 -6.390 -11.01
Garbage_collected -0.024 -4.68 5.642 9.62
Electricity -0.168 -24.66 -3.134 -4.16
Sewerage Network 0.022 6.24 3.495 8.09
Has_Telephone -0.004 -0.42 18.304 15.95
Has_computer 0.294 8.06 46.048 10.50
Adjusted R-squared: 0.4702 0.5520 0.3491y = -0.264x - 6.991
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
End
oge
no
us
Var
iati
on
20
00
-10
(p
erce
nta
ge
po
ints
)
Endogenous 2000
Convergence
67
Another statistics across these series of regressions that is worth looking at is the adjusted
R2 that captures the potential explanatory power of infrastructure in terms of the various
dimensions of social progress presented in the graph below. For example, in the low end,
the gross contribution of this vector on child mortality below one year of age is 13.9%.
On the high end, two thirds of the Human Development Index variation across Brazilian
municipalities is explained solely by this six-fold infrastructure vector.
Source: FGV Social / CPS from the Demographic Census IBGE microdata; Ipea,
UNDP and FJP (2013) and INEP/MEC.
The Human Development Index regression, as its name suggests, is a more encompassing
measure of social progress, including inside by construction the effects of other variables
considered. The table and the graphs below illustrate the correlation between the growth
of the Human Development Index.
# This regression is for the endogenous variable in percentage against its variation in percentage points
## sample for 5010 cities between 2007 and 2015
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Basic Education Index (IDEB) for the 5th grade##
Basic Education Index (IDEB) for the 9th grade##
School Attendance - Children 4-6 years
School Attendance - Children 6-14 years
School Attendance - Children 6-17 years
Life Expectancy
Gini Index
Poverty (Proportion of Poor)#
Child Mortality Under 1 year
Child Mortality Under 5 years
HDI Income Component
HDI Health Component
HDI Educational Component
Human Development Index (HDI)
Per capita Household Income
Per capita GDP
Regressions for Rates of Change across 5500 Municipalities between 2000-2010Explanatory Power of Infrastructure Variables
Gross Contribution Net Contribution
68
The growth regression exercise with infrastructure variables as explanatory variables was
to some extend unsatisfactory when one look at the signs of the infrastructure variables.
We should do tests taking advantage the availability of household surveys microdata.
Before moving to the next step of the analysis it is useful to pose a few additional
questions, namely: Why income convergence between states in Brazil matters? There is a lot
of income inequality within States. Why not looking at overall inequality directly? Should we
invest in poor States, or in poor people anywhere in the country? Should we be looking in
broader terms a social welfare that combines lower overall inequality and higher overall growth?
What type of inequality measure should we use? Overall inequality measures such as Gini,
Atkinson or Theil that gives a lot of weight to the top of the income distribution? Or should we
use instead a poverty measure and derive inequality measures from there? To be sure, what is
the objective function to be pursued more standard Social Welfare Functions, or poverty
directly.
Endogenous Variable: Human Development Index (HDI) Growth 2000-2010
Coefficients Stat t Coefficients Stat t Coefficients Stat t
Intercept -0.154 -71.08 -0.197 -22.98 0.720 108.29
LN (Endogenous) -1.514 -212.73 -1.669 -117.32
Water_network -0.016 -4.40 -0.018 -2.60
Garbage_collected 0.020 5.14 -0.071 -10.10
Electricity -0.026 -4.57 -0.389 -42.93
Sewerage Network -0.019 -6.86 -0.020 -3.86
Has_Telephone 0.101 12.80 -0.233 -16.85
Has_computer 0.237 8.32 -0.234 -4.43
Adjusted R-squared: 0.8916 0.9037 0.6626
y = -1.514x - 0.154
-0.20
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
-0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00
End
oge
no
us
Var
iab
le V
aria
tio
n 2
00
0-1
0
LN (Endogenous) 2000
Convergence
69
Another related question: should we use per capita GDP or incomes that come directly from
household surveys? The former is more related to growth and production considerations which
may be a key intermediary step. While the latter is a closer source of peoples well-being. First,
because the concept itself that is more related to what accrued to people and second because
it allows to calculate the dispersion of income between individuals.
The next step is to construct from household surveys microdata a platform to test the
social impacts or at least the correlations social outcomes with the series of infrastructure
variables proposed. First, we construct from PNAD 2004 and 2015 a series of social
results variables which includes per capita household income (total sources and labor
earnings), years of schooling (for the whole population and for people between 7 and 15
years of age). Imputed rents coming from a hedonic equation and the opportunity time
cost of commuting time evaluated at individual hourly wage rates.
70
8. Distributive Analysis of Infrastructure Social Impacts
We now address possible social impacts of our basic infrastructure items on the
distribution of a vast array of end variables. These social results include per capita total
income, per capita labor earnings, years of schooling, rental value and transportation time
cost. To capture the distributive asymmetry of these impacts we use a quantile regression
approach divided by vintiles. This approach allows us to isolate changes observed along
the distribution of these variables controlled by similar variables used in the other
regressions such as gender, race, city size, unit of federation and second degree
polynomials for age, household size, per capita income9 plus a series of categories for
infrastructure elements discussed above. We opt here to use more detailed categories for
each type of infrastructure. This allow us to recognize finer differences in their potential
impacts. For example, in the case of water services instead of using the binomial having
or having not access to water network we use water sources coming from wells and local
fountain, common in rural areas, and also other alternative water sources to incorporate.
We do not attempt to extract causal relationships here because we do not have
experiments or quasi experiments to warrant for each of the variables analyzed, as in the
previous estimation of income impacts on infrastructure coverage. Nevertheless, these
exercises do provide useful information. In the case of income and rental equations this
type of exercise can help to target socio-economic groups for the selection of beneficiaries
of social programs. We use a traditional mincerian log-linear equation approach to
measure these influences on income. In the case of rental value, we estimate an hedonic
log linear equation incorporating housing characteristics such as number of rooms and
number of bedrooms, the existence and location of bathrooms, type of house (or
apartment), type of construction materials used in walls and ceiling. This is a typical
exercise that allows the estimation of housing wealth, that is the most important physical
wealth component, or alternatively to be incorporated on top of income as imputed rent
values. Similarly one can deduct directly from income the opportunity cost of
transportation of those that go directly to work using the respective hourly-earnings
values. Labor earnings is another key determinant of total income while years of
schooling is the most important determinant of both income variables. In turn, years of
schooling for the 7 to 15 years of age group provides a flow perspective on the
9 Except when we use income as the explained variable.
71
determinants of the stock of the quantity of education. Both equations are regressed in
levels.
We will discuss for each social result variable the regressive, neutral or progressive
pattern of impacts for each infrastructure variable using a graphical perspective and
reporting the results for the 40th and 90th percentiles. Also for the sake of concision, we
present here an analysis of various infrastructure items following an increasing order of
magnitudes around the median of BSW10. Lack of Electricity - The coefficient by those
who use Oil, Kerosene or Gas as sources of light in comparison with those that have
electricity at home as a general rule presents a robust negative sign in all results variables
tested. Total Income plus Imputed Rent Minus Transportation Time Cost – In this
overall welfare measure coefficients are always negative and reaches the bottom at the
60th percentile. The distribution reaches -6% at the 40th percentile and -7.8% at the 90th
percentile. Total (reported) Income - always negative effect rises from -7.5% at the 40th
percentile to -9.4% at the 90th percentile; Labor Earnings – Starts positive but is negative
in 18 out of 20 vintiles. Does not change much when comparing -4.3% at the 40th
percentile to -4.4% at the 90th percentile; Rental Value - always negative effect, bigger
than the previous variables. It falls from -42% at the 40th percentile to -19.3% at the 90th
percentile. Years of Schooling – almost always negative effect rises from -25.3% at the
40th percentile to -38% at the 90th percentile; Years of Schooling (flow) for population
between 7 and 15 years- almost always negative effect. Falls from -46.9% at the 40th
percentile to -25.8% at the 90th percentile11.
Lack of Water - The coefficient of those with no connection to Water Network at home
as a general rule also presents a robust negative sign in all results variables tested. Total
Income plus Imputed Rent Minus Transportation Time Cost – Coefficients are
always negative and reaches the least negative values around the median. The distribution
of coefficients reaches -20% at the 40th percentile to -19.2% at the 90th percentile. Total
Reported Income - always negative effect. Relatively stable around -18% between 40th
10 The reader is invited to analyze the distribution for each type of infrastructure impact of each social
component in the appendix. Including the full specification of the model, the graphical results for other
categories and the full set of equations estimated.
11 Imputed Rent – It is also always negative ranging from -4.91% at the 40th percentile to -6.65% at the 90th percentile. We present the results in the appendix, but we decided to include imputed rent in all analysis since its effects are included in the most general measure used here and some of its effects can be grasped through the rental value hedonic equation above in this paragraph.
72
and 90th percentile, more negative at the extremes of the distribution ; Labor Earnings –
very similar pattern to total income across vintiles, with coefficients 2 to 3 percentage
points more negative; Rental Value - always negative effect. Has a negative trend as we
move to the top of rents distribution little around -25% fluctuates from the 40th percentile
to the 90th percentile. Years of Schooling – somewhat similar pattern to last two items
coefficients fluctuate from -30% at the 40th percentile to -36% at the 90th percentile;
Years of Schooling (flow) for population between 7 and 15 years- effect more negative
at the basis of the distribution. Its negative effect falls from -28.7% at the 40th percentile
to -16% at the 90th percentile;
Lack of Sewerage – Coefficients of those who live in dwellings with Rudimentary
Cesspit compared with those that have a Sewerage Network connection at home presents
a very robust negative signs in all results variables tested except years of schooling for
those at the age corresponding to primary level of education12. Total Income plus
Imputed Rent Minus Transportation Time Cost – The effect increases almost
monotonically in absolute value as we move to the upper tail of the distribution from -
18.3% at the 40th percentile to -24.3% at the 90th percentile. Total Income - always
negative effect with an inverted U-shaped pattern rising in the intermediary interval
between -12.3% at the 40th percentile to -18.2% at the 90th percentile; Labor Earnings –
very similar pattern to total income across vintiles, with similar magnitude of coefficients
2 to 3 percentage points more negative (bigger); Rental Value - always negative effect
with an almost monotonic increase in its negative effect. It falls from -22.6% at the 10th
percentile to -18.6% at the 40th percentile reaching -14.4% at the 90th percentile. Years
of Schooling – usually negative effect but higher at the middle of its distribution.
Negative effect falls from -37% at the 40th percentile to -32.7% at the 90th percentile;
Years of Schooling (flow) for population between 7 and 15 years- small coefficients not
always negative effect. It changes from -1.52% at the 40th percentile to -0.05% at the 90th
percentile;
Communication - The impact coefficient of those who are in dwellings with telephone
or cell phones for at least one of the household members compared to the rest of the
population without this device. Note that we are looking now those who have access
compared with those who have not so all the signs in the impact analysis of infrastructure
12 This pattern replicates itself for other types of sewerage especially those not connected to any network, especially those related to sewage directed to natural water deposits.
73
work the other way around. Most of the effect is due to cell phone possession that became
much more diffused than landline phone. As opposed to the internet, the total income
effect is higher than the labor earnings effect but both remain higher than the rental value
effect. The cell phone effect is relatively higher on the basis of the distribution than
internet access. The statistics organized by type of social outcome show that: as a general
rule, communication coefficients present a robust positive sign in all results variables
tested. Total Income plus Imputed Rent Minus Transportation Time Cost – Overall
effects increases from 34.7% at the 40th percentile to 43.4% at the 90th percentile. Total
Income - always positive effect following an U shaped pattern with bigger effects on the
extremes of the income distribution. This suggests that the bottom also benefits a lot from
cell phone contrary to what happened with internet access. In the 5th percentile the
coefficient is 36.4% reaches 32% at the 40th percentile then rises back to 36.5% at the 90th
percentile; Labor Earnings – always positive effect rising almost monotonically
suggesting that the bottom part of the distribution does not benefit as much as in the total
income which may suggest that is not a labor related issue. The coefficient rises when
comparing 30.6% at the 40th percentile to 36% at the 90th percentile; Rental Value -
always positive effect with a declining trend especially as we move from the 5th to the
20th percentile stabilizing between 12% and 11% from this point onwards. The respective
coefficient is much smaller than the income and labor earnings coefficients. Years of
Schooling always positive effect following an inverted U shaped pattern with smaller
effects on the extremes falling from 149% at the 40th percentile to -124% at the 90th
percentile; Years of Schooling (flow) for population between 7 and 15 years- With the
exception of zeros in the upper half of the distribution falls from 29.4% at the 40th
percentile to 27.9% at the 90th percentile.
Internet - The impact coefficient of individuals in dwellings with internet access
compared with those without it presents a robust positive and high sign in all results
variables tested. The coefficients presents a positive trend as we move towards the top of
each distribution, suggesting at face value that those at the top benefit relatively more
from internet access. The income variables related coefficients increase along each
particular concept. As a consequence, the diffusion of internet should lead to a divergence
in these different social outcomes. Total Income plus Imputed Rent Minus
Transportation Time Cost – effects increases from 58.4% at the 40th percentile to
82.7% at the 90th percentile. Total Income – always positive effect rising almost
74
monotonically. It changes from 52.8 % at the 40th percentile to 68.8% at the 90th
percentile; Labor Earnings – Also always positive effect rising almost monotonically
along the distribution. Effect very much alike the previous total income effect but a little
steeper. Magnitude 2 to 3 percentage points lower in bottom percentiles and 2 to 3 points
higher in top percentiles. It changes from 49.2% at the 40th percentile to 71.2% at the 90th
percentile; Rental Value - always positive effect also with an upward trend as we move
towards the top of the distribution but its magnitude is smaller than the income variable.
Rises from 11.5% at the 40th percentile to 12.8% at the 90th percentile. Years of Schooling
–always positive and substantive effects, more pronounced in the core of the formal
human capital distribution. This effect falls from 159% at the 40th percentile to 107% at
the 90th percentile; Years of Schooling (flow) for population between 7 and 15 years-
positive in most parts of the distribution with a few zeros. It rises from 14.4% at the 40th
percentile to 21.3% at the 90th percentile;
Commuting Time evaluated at hourly-wage rate – It works as an approximation to
transportation cost in urban areas. It is included in the broader welfare measure. We just
check whether it has increased from 2004 in 2015 and its distributive change pattern. The
5% poorest had the highest increase of 41.1% that tended to decrease reaching 33.6% at
the 40th percentile with some stability reaching to 32.3% at the 90th percentile, then rising
to 35.4% in the top vintile;
75
9. Ranking Infrastructure Direct Social Impacts & their Externalities
Stepwise Models -Instead of imposing a particular model of analysis, we implement here
a stepwise variable selection procedure to determine which socio-economic and
infrastructure related variables are more statistically important to explain each social
outcome variable seen in section 8. We also include poverty in this analysis using a
binomial logistic regression, the remaining variables we apply an OLS log-linear
minceriana regression. Given the results of the quantile regressions where each category
of each infrastructure variable was tested, we use here a twofold division of the most
relevant category for each variable. In the selection process we included variables that
capture externality effects from infrastructure. This is done by including in the regressions
the mean of these variables across geographic areas. The idea is to see how much a
subdivision of the 27 units of the federation into three or four areas each namely rural,
urban non metropolitan or capital of the state. In the case of the states that include one of
the 11 major Brazilian metropolitan cities we include a finer division between capital and
suburbs for these metro regions. Given the difference in economies or diseconomies of
scale between cities sizes. After the variable selection process, we discarded externality
related variables with signs that are in disagreement with the expected sign provided by
theory. The idea is that beyond individual impacts at the household level, what our
neighbors and other community members have in terms of infrastructure use may also
affect ours respective social outcomes. For example, if there is a widespread diffusion of
landline or cell phones in my region of residence the value of my phone line increases
due to network scales, given the fixed cost of intercity connections. Following a different
strand the effects of electricity access at the community level may also improve my social
outcomes through better work opportunities or school or health services. Transportation
use on the other extreme imply a common good congestion problem where the excessive
use of infrastructure generates a negative externality on all users. The order of variable
selection is indicative of the relevance reached by each explanatory variable.
Poverty - In the case of the proportion of the poor, the six infrastructure variables are
significant in descending order: communication, internet, transportation, water, electricity
and sewerage. Two of the externality related variables also presented statistically
significant impacts, namely mean transportation time and mean electricity coverage. The
respective regression coefficients can be found in the appendix.
76
Mean Broader Welfare - For broader social measure mean - that includes besides total
income sources from PNAD survey, imputed rents from housing minus opportunity time
cost of commuting at individual level – the results are similar to poverty, with ICTs and
transportation time presenting highest significance. Externalities with respect to
electricity and transportation time are also included in the final model. Internet related
infrastructure at the regional level does not show any geographical externality, which is
somewhat expected since the use of the so-called world wide web allows to overcome
these location barriers. One difference is that externality of communications appears here
as one of the top variables.
Externalities and other social outcomes - The existence of intercity and inter-state costs
makes the case for stronger externality at the local level. If we look at total per capita
income as well as labor earnings they both are show externality effects in the same fields
of phone communications and transportation. In contrast, completed years of schooling
are affected by internet related infrastructure. This may be a proxy for the effects of the
digital age in schools, libraries and so on. When we restrict this variable to school age
between 7 and 15 years of age, the main externality is yield by electricity. Programs like
Light in Schools (Luz na Escola) and Light for Everybody (Luz para Todos) attempt to
explore this effect. Imputed rents indicate that housing values are also affected by phone
communications and transportation costs, especially the former that occupies the top
position among all explanatory variables.
77
10. School Performance and Infrastructure
Using the microdata of the Basic Education Evaluation System (SAEB/MEC) of 2003
and 2015, we estimated the impact of infrastructure variables in school proficiency and
grade repetition. This is done combining the objective infrastructure coverage
information at students home and at school with the perceived quality of infrastructure
services in school and running OLS regressions explaining proficiency tests outcomes in
levels. We focused especially in kids in the fifth grade, once this group represents the
youngest group evaluated and the next generation within Brazilian workforce. Results for
the fifth grade in Mathematics and Portuguese language reveal an improvement of the
educational system during this period, with a mean increase of almost 35 and 31 points,
respectively. Multivariate results do not allow us to reject the hypothesis that investment
in public infrastructure services is more important for proficiency improvement than
typical physical investment in school buildings, once good electricity and water
installations had a higher impact than the conservation status of classrooms and
bathrooms. Robustness tests were made with the math exams of students in the ninth
grade and in the last year of high school.
Between 2003 and 2015, Brazilian students in the fifth grade improved almost 25% their
proficiency in math tests, going from a mean of 177 points up to a mean of 219 points, an
improvement of 42 points. In both years, students with bathroom at home had a better
performance than those who had not. The same pattern was observed for students with
computer at home. In terms of schools infrastructure in terms of attributes like electricity,
water, illuminated and well-made classrooms and bathrooms, students enrolled in schools
with good infrastructure had higher proficiency. Nevertheless, what can we say about the
impact of the verified coverage expansion of these infrastructure attributes on the upward
trend in math proficiency? In other words, have they contributed for the recent proficiency
improvement?
Multivariate results for the math proficiency of the fifth grade students were controlled
for year (2003 and 2015), student characteristics (sex and color), household assets
infrastructure (existence of bathroom in student house and existence of computer in
student house), school characteristics (if school is private or public and rural or urban)
and school assets infrastructure (has good illumination and well-made classrooms, has
good bathrooms, water installations and electricity). Students with the same household
and school characteristics had an improvement of almost 35 points in 2015 compared
78
with 2003, which represents a progress of the quality of education13. We observed the
same pattern for similar students that differed only in terms of infrastructure coverage,
whether at home or school, as the graph below shows. Those with access to good
installations of electricity and water in school had a math proficiency, in average, 7 and
6 points higher, respectively. It is interesting to notice that classrooms walls in good
status, our proxy for well-made classrooms, showed little importance for the outcome,
suggesting at face value that investment in public infrastructure services that is connected
outside schools was more important for proficiency improvement than typical private
investment in buildings. However, the quality of bathrooms seemed important, once
students with access to good bathrooms had proficiency 9 points higher. Robustness tests
for Mathematics in the ninth grade and the last year of high school, besides Portuguese
exams for the 5th grade, generated interesting results. Both ninth grade and last year of
high school presented similar results in math proficiency for household and school assets
infrastructure. The difference were the neutrality of good electricity installations and the
kickback of mean proficiency measured by the dummy for 2015 in the last year of high
school estimations. On the other hand, Portuguese language for the fifth grade had a
similar evolution process. The average proficiency advanced 22.5%, going from 170 up
to 207 points, while students with household assets infrastructure and school assets
infrastructure had higher scores. The controlled multivariate tests showed that similar
students in schools with good private and public infrastructure were better ranked.
Electricity (10 points), bathroom (8 points) and water (5 points) were the main
infrastructures assets of impact. The improved quality hypothesis also remains, with a 30
points average for students in 2015 than in 2003, given their controlled characteristics.
13 Taking into account the hypothesis that our coefficient is not being influenced by omitted variables or any kind of bias.
79
The difference-in-difference method provides a dynamic analysis of the infrastructure
contribution, once it compares the difference in proficiency between students with access
to an infrastructure asset in 2015 and 2003 with the difference between the group of
students marginalized in terms of these assets in both years. Controlling for home and
school attributes, students with access to good electricity and water installations in school,
compared with those without that, had an average proficiency improvement of 27 and 11
points, respectively, between 2003 and 2015. In the other hand, at the same period,
proficiency of students with access to good bathrooms and classrooms in school had no
statistical difference than of students enrolled in more precarious schools. Therefore, the
diff-in-diff test corroborates the main role of public infrastructure in the recent upward
movement of school proficiency in the fifth grade. Robustness tests for math exams for
the ninth grade and the last year of high school showed no statistical significance for good
electricity installations diff-in-diff coefficient, however, most water interaction
coefficients were with switched sign, especially for the last year of high school.
The diff-in-diff method for proficiency in Portuguese language for the fifth grade,
notwithstanding, presented new features. While good quality of electricity installations at
school remained the main infrastructure attribute, with 28.6 points of difference in favor
or those with access between both years, coefficient of good quality of water installations
#Controlled for household assets infrastructure, student general characteristics, school assets infrastructure and year of the survey
## All coefficients significative at 99%
Source: FGV Social/CPS using SAEB/IBGE microdata
-15.0
-29.4
-3.3
9.16.4 7.2
34.6
-15.2
-31.7
-2.1
7.95.3
10.4
30.6
-35.0
-25.0
-15.0
-5.0
5.0
15.0
25.0
35.0
No ComputerHome
No BathroomHome
Badly IlluminatedSchool
Bathroom School Water School Electricity School Proficiency_Diff(2015-2003)
Proficiency Impact for Private and Public Infrastructure Assets Controlled# Multivariate Tests for the 5th grade
MATH PORT
80
and bathroom, were not statistical significant14. Well-made classrooms coefficients
neither.
We also applied a process of variable selection using a stepwise statistical procedure. This
program evaluates all the variables pre-selected to the model and rank them by better
adjustment with the variable of interest. In both models for fifth grade mathematics
proficiency (with and without interaction variables), the champion and runner-up
variables were “computer at home” and “color” of the student. “Bathroom at home” and
“local of the school” (urban or rural) were in the third and fourth positions for the model
without interaction, respectively. Both variables lost one position for the interactive
variable “Bathroom at school*Dummy 2015”, that was in third place in the model with
interactions. Water and Electricity installations were in top 10 in both models. Eighth and
ninth positions, respectively, in the model without interaction, and ninth and tenth
positions, respectively, in the model with interaction. In both models for fifth grade
Portuguese language proficiency, “computer at home”, “sex”, “color”, “bathroom at
home”, “water installations” and “electricity installations” occupied the 1st,2nd,3rd,4th,8th
and 9th places, respectively. “Computer at home” was also the champion in both models
for the ninth grade, whether for math or Portuguese language. The type of school (public
or private) was the most important variable in the models for the last year of high school.
Grade Repetition and Infrastructure – To make a parallel of the present infrastructure
analysis with changes of the so-called IDEB (Basic Education Development Index), we
14 Surprisingly, regular quality bathroom had an impact of 7.7 points.
#Controlled for household assets infrastructure, student general characteristics, school assets infrastructure and year of the survey
## All coefficients significative at 99%
Source: FGV Social/CPS using SAEB/IBGE microdata
0 0
10.95
27.48
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
Well-MadeClass_School*D2015 Bathroom_School*D2015 Water_School*D2015 Electricity_School*D2015
Diff-in-Diff Proficiency Impact for Infrastructure School AssetsControlled# Multivariate Tests for the 5th grade
MATH PORT
81
use the question of SAEB on grade repetition to proxy flow variables in IDEB. IDEB is
a synthetic indicator of education quality based on the academic passing rate and the
results of proficiency exams (as SAEB and Prova Brasil) for each municipality and school
in the country. As we have seem in section 7, among many different social outcomes,
IDEB across the Brazilian municipalities converged at a higher speed in the last decade
in the presence of infrastructure variables, meaning the municipalities with lower initial
educational performance grew faster than the higher ones and this speed was influenced
by infrastructure. In this section, we are attempting to mimic the flow of students captured
in IDEB using the SAEB data. The main questions is: Do infrastructure variables affect
grade repetition? To answer this question we generated logistic regressions using a
dummy for students that have repeated at least once. As in the previous section, our model
controls for year (2003 and 2015), student characteristics (sex and color), household
assets infrastructure (existence of bathroom in student house and existence of computer
in student house), school characteristics (if school is private or public and rural or urban)
and school assets infrastructure (has good illumination and well-made classrooms, has
good bathrooms, water installations and electricity).
Results for the non-interactive model showed statistical significant coefficients for
household assets, with 12% and 37% more chances for repetition for students without
computer and bathroom at home, respectively. However, the quality of classrooms
physical structure and illumination apparently did not affect grade repetition. The only
school private infrastructure with positive impact was the quality of bathrooms, with 24%
less chances for repetition for students with good bathrooms in their schools. While water
installations did not improve school flow (with more chances of repetition for all
coefficients), students in schools with good electricity installations had 9% less chances
of repeating their grade. The time variation, measured by the dummy for 2015, suggested
a huge advancement in the quality of education in this grade, with 95% less chances of
repetition for students in 2015 in comparison with peers with the same scholar and home
characteristics and infrastructure in 2003. Robustness tests for the ninth grade and the last
year of high school corroborated the importance for household assets in reducing
repetition probability. In the other hand, good water installations at school had positive
impact in school flow in both groups (12% and 25% less chances of repetition,
respectively), while good electricity installations and bathroom at school were statistical
neutral for the last year of high school and positive for the ninth grade (17% less chances
of repetition for both infrastructure variables). The dummy for 2015 also showed a
82
progress in the quality of education for peers in the ninth grade and the last year of high
school (29% and 40% less chances of repetition, respectively).
The difference-in-difference method for the fifth grade measured by the interaction of
school infrastructure variables with the dummy for 2015 captured a positive impact of
51% less chances of repetition for students in schools with good bathrooms, compared
with peers without this infrastructure variable, during this period. However, the quality
of classroom physical structure, water installations and electricity had no statistical
impact on repetition of the fifth grade between both years. Robustness tests for the ninth
grade and the last year of the high school presented different features. In none of them
the interaction between the quality of bathroom and the dummy for 2015 were statistical
significant. For the ninth grade, the only coefficients with marginal statistical significance
were about physical structure of classrooms. However, we cannot infer that well-made
classrooms are better than poor-made ones because both have similar impacts compared
with peers without walls in their classrooms. In turn, the only statistical significant
coefficient for the last year of the high school were the interactive variable for bad quality
water installations and the dummy for 2015, suggesting a higher importance for the
coverage than the quality of this public service.
83
11. Conclusions and Prescriptions
This paper has provided an empirical analysis on the access to public services
infrastructure in order to base prescriptions for improvement policies. The final objective
of this work is to create a basic infrastructure of knowledge to guide the use of the new
generation of programs in the universalization of public utility services.
A first contribution of this work was to analyze in a comparative way the coverage of
these surveys with different databases, including information provided by service
providers and even School Census, in order to more critically analyze their evolution and
create monitoring systems. The household survey approach is also particularly useful
because it allows to study a vast array of social consequences derived from infrastructure
expansion. Attributes of the various public services through household surveys, such as
spatial coverage, perceived quality, expenditures and delay of accounts. We compared
also the Perceived Quality and Priorities given to infrastructure sectors - In general,
the quality of services associated with water enjoys lower perceived quality than that of
public services such as electricity and garbage collection. Besides the subjective quality,
an analysis of the priorities of the Brazilian population that out of 16 new Sustainable
Development Goals (SDGs), infrastructure variables stay in the following positions:
Transportation (7th); Water and Sanitation (9th); Electricity (13th) and ICTs (16th).
Access to infrastructure services has increased significantly over the past decade.
This is mainly due to lagged effects of the privatization programs of the 1990s (especially
in telecommunications), the adoption of public programs aimed at expanding coverage in
remote areas (especially in electricity due to the “Luz Para Todos” program) and the
demand effect from the combination of faster household income growth and falling
inequality that lasted until 2014. Using household level data on coverage of infrastructure
services, the service that had the highest increase in access between 2004 and 2015 was
ICT. The past 10 years has seen an explosion in the use of mobile telephones. In 2004,
around 85 million people had mobile phones at home, and in 2015 the number increased
to 186 million – an increase of 101 million users. During the same period, home internet
coverage was extended to an additional 64 million Brazilians. Despite its rapid growth,
internet service is the infrastructure service that presents the lowest level of access (42.5
percent) when compared to other services. On the other extreme is electricity, with an
access level of 99.7 percent. Access to potable water has an intermediate rate of 83.6
percent, but significantly more than sewage services, at 56.9 percent.
84
Coverage of infrastructure services in rural areas has expanded but the sharp divide
between rural and urban coverage within the country persists. Only in sanitation has
rural coverage not changed much. However access gaps between rural and urban areas
remain high. While rural areas represent around 14 percent of the Brazilian population in
2015, only 4 percent of this population has access to sewerage services with only a third
having access to the water system. In urban areas, where most of the population lives, the
rate of access to the water system is about 90 percent while access to sewerage services
is about 80 percent. The pattern of low rates of access in rural areas and high rates of
access in urban areas is evident in all infrastructure services with the exception of
electricity where access rates have converged.
As a product of the demographic transition household size had fallen 1.43% per year
bigger than the 0.8% per year of total population size growth rate. This means that the
supply of infrastructure has to increase not only because of the existing infrastructure
deficit and population growth but also as a response of the household size reduction.
Infrastructure access reflects and reinforces Brazil’s poverty profile and extreme
high income inequality. Access rates among the poor have been improving in the last
decade but coverage remains much higher among wealthier groups. Sewerage, water and
internet tend to be the most unequally distributed services across income groups. In 2015,
less than half of the poorest segment of the population had access to sanitation facilities,
compared with 80 percent of the richest.
Income Causality - How much access to public infrastructure is related to exogenous
increase of income. through the Bolsa Familia program. The results are a relative
improvement for all infrastructure items, except sewerage. This lack of sensibility may
be due to the predominance of externalities in the supply of sewerage where individual
or private returns to sewerage connection benefits mostly others.
Infrastructure Convergence - Multivariate exercises revealed that keeping socio-
demographic structure constant, the highest temporal change between 2004 and 2015 was
observed in electricity, internet and cell phone. The lowest expansion was found in water
and sewerage. State level dummies has taught us that São Paulo presents the best
infrastructure across Brazilian States. While interactions between State and year dummies
has shown that the differential between different states and São Paulo tended to fall. This
85
shows a clear convergence trend of infrastructure between Brazilian States even if we net
out the effects of income, education and other variables during this period.
Social Convergence - We implemented a standard convergence analysis running
regressions across 5500 Brazilian municipalities of growth of each variable against the
natural logarithm of initial value comparing the results with and without infrastructure
variables. The set of variables tested includes per capita GDP, per capita household
income, the Human Development Index, its 3 components plus a series of related
variables such as poverty and inequality, life expectancy, child mortality, school
attendance for various age brackets and the Basic Education Development Index (IDEB)
which includes the results of proficiency exams. For 16 out of the 17 endogenous
variables tested the speed of convergence is higher at face value with the set of
infrastructure variables than the model without infrastructure. The exception is per capita
GDP where the size of the lagged endogenous variable coefficient decreases in absolute
terms when we include the infrastructure parameters. The gross explanatory power of
infrastructure in terms of the various dimensions of social ranges from 13.9% on child
mortality to 66% for the Human Development Index.
Distributive Impacts - Quantile regressions based platform infrastructure variables
social impacts along the distribution of different outcomes which includes per capita
household income (total sources and labor earnings), years of schooling (for the whole
population and for people between 7 and 15 years of age). Imputed rents coming from a
hedonic equation and the opportunity time cost of commuting time evaluated at individual
hourly wage rates. For the sake of concision, we emphasize here the potential distributive
impacts on the broader social measure that includes total reported income plus imputed
rent minus commuting costs. We present here following an increasing order of
magnitudes around the median for various infrastructure items. Lack of Electricity
coefficients reaches -6% at the 40th percentile and -7.8% at the 90th percentile. Lack of
Water coefficients reaches -20% at the 40th percentile to -19.2% at the 90th percentile.
Lack of Sewerage coefficients from -18,3% at the 40th percentile to -24.3% at the 90th
percentile. Communication - The coefficient of those who are in dwellings with
telephone or cellphones increases from 34.7% at the 40th percentile to 43.4% at the 90th
percentile. Internet - The coefficients effects increases from 58.4% at the 40th percentile
to 82.7% at the 90th percentile presents a positive trend as we move towards the top of
86
each distribution, suggesting at face value that those at the top benefit relatively more
from internet access. As a consequence, the diffusion of internet should lead to a
divergence in these different social outcomes.
Infrastructure Externalities - We implemented a stepwise variable selection procedure
to determine which socio-economic and infrastructure related variables are more
statistically important to explain each social outcome variable seen above. In the selection
process we included externality effects from infrastructure. Poverty - In the case of the
proportion of the poor the six infrastructure variables are significant in descending order:
communication, internet, transportation, water, electricity and sewerage. - Broader social
measure mean that includes besides total income sources from PNAD, imputed rents from
housing less opportunity time cost of commuting– the results are similar to poverty. On
both social outcomes. two of the externality related variables presented statistically
significant impacts namely mean transportation time and mean electricity coverage.
Electricity access at the community level may improve individual social outcomes
through better work opportunities or school or health services. Transportation use on the
other extreme imply a common good congestion problem where the excessive use of
infrastructure generates a negative externality on all users. Externality of communications
appears here as one of the top variables but only in mean broader social welfare
measure. The existence of intercity and inter-state extra calling costs makes the case for
externality for phones at the local level. Externalities with respect to electricity and
transportation time are also included in the final model. Internet related infrastructure at
the regional level does not show any geographical externality which is somewhat
expected since the use of the so-called world wide web allows to overcome these location
barriers.
Education and Infrastructure - The interaction between infrastructure related physical
capital and human capital occupies a central role in the analysis. School quality
convergence - among many different social outcomes, IDEB across the Brazilian
municipalities converged at a higher speed in the last decade in the presence of
infrastructure variables, meaning the municipalities with lower initial educational
performance grew faster than the higher ones and this speed was influenced by
infrastructure. To make a parallel of the present infrastructure analysis with changes of
the so-called IDEB (Basic Education Development Index), we use the question of SAEB
on grade repetition to proxy flow variables in IDEB. Grade Repetition - Results showed
87
statistical significant coefficients for household assets, with 12% and 37% more chances
for repetition for students without computer and bathroom at home, respectively. While
water installations did not improve school flow, students in schools with good electricity
installations had 9% less chances of repeating their grade. School Proficiency
SAEB/MEC tests were also used - We cannot reject the hypothesis that home
infrastructure coverage and investment in public infrastructure services is more important
for proficiency improvement than typical physical investment in school buildings.
Policy Prescription - After comparative empirical analysis of the various public services
in different databases, we return to the analysis for the sanitation sector. For three reasons:
the first is the evidence of lower coverage, poorer quality and stagnation of sanitation
coverage in the country compared to other public services. Secondly, the deleterious
impacts of sanitation on all dimensions of human development, by the health of people
in general and children in particular. Finally, in addition to the importance of sanitation,
we need to take into account the specificities of the sector's enormous challenges, such as
the lower visibility of its impacts by the population and associated coordination problems.
Despite the existence of large investments in public infrastructure, such as the announced
Growth Acceleration Program (PAC), the new Basic Sanitation Law and a certain
mobilization of public opinion, the incentive structure for the provision of public services
has not helped.
The results suggest that the difficulty of sanitation vis-a-vis other public services is not
only a lack of income. The lack of light or water is obvious to the ordinary citizen in their
daily life, however the lack of sanitation is not. It is a problem of others. In this context,
the individual ideal is for others to collect their respective sewage. Now if everyone thinks
so, we all end up living next to open ditches. The collection of sewage is not perceived as
an individual gain, it is therefore a challenge of collective action. Now how to do it? That
is the question. A response should be "Bolsa Saneamento", which refers to the use of the
Bolsa Família program structure for the provision of incentives to consumers and
companies in expanding coverage of sewage collection and its treatment.
We focus on the possibilities offered by Bolsa Família, whether to test the impact of
income on access to services, or as a platform for granting subsidies. The marked
88
expansion of the program between 2004 and 2006 served initially as an experiment about
the impacts of the income increase associated with policies to combat poverty on the
coverage of public services. We analyze how much the income increase of this population
is related to the increase of their access to public services. The results show that, in the
controlled analysis, cellular was the only service that grew in the period. We combined
the variables year and program eligibility to measure whether with the income gain the
access of the population of low income grew more than the others. The results are positive
in the case of cellphone access and access to the general water network. However, in
sanitation there was no statistically significant improvement over the other group. The
higher income provided by the Bolsa Família program did not affect the access to
sanitation of the population eligible for the program. This may be due to the operation of
externalities. This point will be analyzed from an empirical perspective in section 9.
If the Bolsa Família program itself was not a sufficient condition to lead to the provision
of sanitation to the poor segments, it serves as a platform for access to the poor through
the single social cadaster used in its operation. As a central policy prescription, we have
the use of the Bolsa Família structure. This possibility is through the availability of the
Single Social Registry (CadÚnico) associated with the operation of the Bolsa Família
program (PBF). The CadÚnico presents the people's financial address associated with the
program's payment card possession while allowing infrastructure programs to connect
with the poorest. In that way, companies can receive incentives for network extension
focused on the poor or a direct subsidy to the value of incentives. In particular, the
association of OBA (Output Based Aid) incentive schemes with Bolsa Família, the
country's main policy to combat poverty, is a privileged way to provide incentives for
public services to reach the poor.
89
Bibliography on ICTs Infrastructure Sectors
AFONSO, C. A. Internet no Brasil: o acesso para todos é possível?. Policy Paper nº
26, Friedrich Ebert Stiftung, 2000. 37 p.
ANATEL. Participação do Mercado por UF. 2011. Disponível em:
http://sistemas.anatel.gov.br/SMP/
ANDERSON, C. The long tail. Londres, Random House Business Books. 2006
ASHTA, A. and ASSADI, D. Do Social Cause and Social Technology Meet? Impact
of Web 2.0 Technologies on Peer-to-Peer Lending Transactions. Cahiers du CEREN,
v. 29, pp. 177-192, 2009.
BUSINESS WEEK. Is the Digital Divide a Problem or an Opportunity? Special
Advertising Section, 18th December 2000.
CANELA, G.; NASCIMENTO, S. Acesso à informação e controle social das políticas
públicas. Brasília: ANDI e Artigo 19, 2009.
CASTEL, R., La inseguridad social ¿Qué ES estar protegido?. Buenos Aires,
Manantial. 2002.
CGI.br. Pesquisa sobre o uso das tecnologias da informação e comunicação no Brasil:
TIC Domicílios e TIC.
CHONG, A., MICCO, A. The Internet and the Ability to Innovate in Latin America.
Working Paper 464, Inter-American Development Bank, 2002.
CHRISTENSEN, C. M. The Ongoing Process of Building a Theory of Disruption.
Journal of Product Innovation Management, 23(5), p. 39-55, 2005.
COMITÊ PARA DEMOCRATIZAÇÃO DA INFORMÁTICA (CDI) e SUN
MICROSYSTEMS. Global Digital Divide Initiative: Steering Comitee on Education.
World Economic Forum, mimeo, 2002
COMSCORE. The 2010 Europe Digital Year in Review. Disponível em:
http://www.comscore.com/Press_Events/Press_Releases/2011/2/comScore_Releases_T
he_2010_Europe_Digital_Year_in_Review
DIGNAN, Larry. Google Makes Waves and May Have Solved the Data Center
Conundrum. 2010. Disponível em: http://www.zdnet.com/blog/btl/google-makes-waves-
and-may-have-solved-the-data-center-conundrum/9937
DUARTE, A. Entrevista com Manuel Castells: “Se um país não quer mudar, não é a
Internet que irá mudá-lo”, diz sociólogo espanhol. Folha de S. Paulo, 21 de setembro
de 2010. Disponível em:
http://bit.ly/9KL95Q.ec.europa.eu/environment/waste/weee/index_en.html
EINHORN, Michael A. "Biases in Optimal Pricing with Network Externalities.” In:
Review of Industrial Organization 8: 741-746, 1993.
90
Empresas 2009 / Survey on the Use of Information and Communication Technologies
in Brazil: ICT Households and ICT Enterprises 2009. Coord. Alexandre F. Barbosa.
Trad. Karen Brito. São Paulo: Comitê Gestor da Internet no Brasil, 2010.
EPA. It’s Easy Being Green: A Guide to Planning and Conducting Environmentally
Aware Meetings and Events. Washington: Environmental Protection Agency, 1996.
EUROPEAN COMMISSION ENVIRONMENT. Recast of the WE Directive. 2011.
EUROPEAN INNOVATION SCOREBOARD. Global Innovation Scoreboard 2008:
The Dynamics of Innovative Performances of Countries. 2009.
EUROSTAT. Eurostat Model Questionnaire for the Community Survey on ICT
Usage and e-Commerce in Enterprises. 2010.
FEDERAÇÃO DAS INDÚSTRIAS DO ESTADO DO CEARÁ (FIEC). Sobre o lixo
tecnológico. 2010. Disponível em:
http://www.sfiec.org.br/iel/bolsaderesiduos/Artigos/Artigo_Lixo_tecnologico.pdf
FRAUNHOFER INSTITUTE. Thin Client 2011: Ecological and Economical Aspects
of Virtual Desktops. 2011. Estudo técnico. Disponível em:
http://www.igel.com/fileadmin/user/upload/documents/PDF_files/White_Paper_EN/thi
nclients2011-en.pdf
FRIEDMAN, T. L. O Mundo é plano. Uma Breve História do Século XXI. Rio de
Janeiro: Objetiva, 2005. 640p.
GOMES, E. Exclusão Digital: um problema tecnológico ou social?. Instituto de
Estudos do Trabalho e Sociedade, Rio de Janeiro: Trabalho e Sociedade, ano 2, número
especial, Dezembro de 2002.
GUZZI, Drica. Web e participação: a democracia no século XXI. 1ª Ed. São Paulo:
Editora Senac São Paulo, 2010. 160p.
INTERNATIONAL TELECOMMUNICATION UNION (ITU). Measuring the
Information Society - The ICT Development Index, 2010 Edition.
INTERNATIONAL TELECOMMUNICATION UNION (ITU). Challenge to the
Network. Internet for Development. Executive Summary, ITU, Geneva, 1999.
KASERMAN, D. and MAYO, J. “Cross-Subsidies in Telecommunications:
Roadblocks on the Road to More Intelligent Telephone Pricing.” In: Yale Journal of
Regulation 11(1), 1994.
KIEWIET, D. R.; MCCUBBINS, M. D. Delegation and Agency Problems. In: The
Logic of Delegation: Congressional Parties and the Appropriations Process. Chicago:
University of Chicago Press, 1991.
KIRKPATRICK, D. O efeito Facebook: os bastidores da história da empresa que
conecta o mundo. 1ª edição. Rio de Janeiro: Intrínseca, 2011. 392p.
91
KRUEGER, Alan B., How computers have changed the wage structure: evidence
from microdata, 1984-89. Working Paper nº 3858, National Bureau of Economic
Research. Outubro de 1991.
LAFFONT, Jean-Jacques & TIROLE, J. “Creating Competition through
Interconnection.” In: Journal of Regulatory Economics 10: 227-256, 1996. 296pp.
LANJOUW, P. and RAVALLION, M. "Poverty and Household Size". In: Economic Journal, 105 e 433, 1995.
MARTINS, A., AFONSO, C. A., ASSUMPÇÃO, R. e BARCELLOS, S. Oficina para a
Inclusão Digital. Plenária Final, Centro de Convenções Ulysses Guimarães, Brasília,
Maio de 2001.
MILLER, Rich. IDS Readies Data Centers on Ships. 2010. Disponível em:
http://www.datacenterknowledge.com/archives/2010/08/09/ids-readies-data-centers-on-
ships/.
MINISTÉRIO DO PLANEJAMENTO, ORÇAMENTO E GESTÃO. 2 Anos de
Governo Eletrônico - Balanço de Realizações e Desafios Futuros. Casa Civil da
Presidência da República, Secretaria Executiva, Brasília, Dezembro de 2002.
MOLINA, A. “The Digital Divide: The Need for a Social Movement". The University
of Edinburgh.
MORANDI, L. Estimação da Riqueza Interna Tangível e Reproduzível - Brasil
1970/95. Universidade Federal Fluminense, mimeo, tese de mestrado, 1997.
MUSTONEN, J. (org.). The World’s First Freedom of Information Act: Anders
Chydenius’ Legacy Today. Kokkola: Anders Chydenius Foundation, 2006.
NERI, M. C. A Nova Classe Média: O Lado Brilhante da Base da Pirâmide, Editora
Saraiva. São Paulo, 2011. 312 p.
NERI, M “O Início, o Fim e o Meio Digital”, mimeo, Rio de Janeiro: FGV/CPS,
Fundação Telefônica/Vivo, 2012.
______. “Inclusão Digital e Educação”. In: BAYMA, Fátima. (Org.). Educação
Corporativa: Desenvolvendo e Gerenciando Competências. 1ª edição. Editora Prentice-
Hall, Rio de Janeiro, 2004, v. 1, p. 220-225.
______. “Inclusão Digital e Redistribuição Privada”. In: Oliveira, Fátima Bayma. de.
(Org.). Seminário Nacional de Tecnologia da Informação e da Comunicação aplicada à
Gestão dos Negócios e Programas Sociais. Rio de Janeiro, 2005. p. 215-230.
______. Mapa da Exclusão Digital na Bahia e desenvolvimento dos Sistemas de
Avaliação de Impacto e de Monitoramento do Programa Identidade Digital.
PESQUISA. Rio de Janeiro: CPS/IBRE/FGV, 2006.
______. O Mapa da Exclusão Digital. Revista Conjuntura Econômica. Rio de Janeiro,
92
p.70 - 73, 2003.
______. O Paradoxo da Evasão e as Motivações dos Sem-escola. In: Educação Básica
no Brasil: Construindo o País do Futuro. 1ª ed. Rio de Janeiro: Campus-Elservier, 2009,
v.1, p. 25-50.
______. O peixe, a vara e a rede de computadores. Revista Conjuntura Econômica. Rio
de Janeiro, v.60, p.41 - 43, 2006.
.
______. "Políticas estruturais de combate à pobreza no Brasil". In: Henriques,
Ricardo (orgs), Desigualdade e pobreza no Brasil, IPEA, Rio de Janeiro, 2000.
______. Retratos da Deficiência no Brasil. Rio de Janeiro: Fundação Getulio Vargas,
2003, v.1. p. 2004.
NERI, M. C.; AMADEO, E. J. E CARVALHO, A.P. "Assets, Markets and Poverty in
Brazil", in Portrait of the Poor – An Assets-Based Approach Orazio Attanasio e Miguel
Székely (org.). IDB, Washington, pp 85-112, 2001.
NERI, M. C., BUCHMANN, G. From Dakar to Brasilia: monitoring UNESCO’s
education goals. Prospects (Paris). , v.38, p.415 - 415, 2009.
NERI, M. C.; MELO, L. C. C. de; NERI, A.L ; CORSI, A. P.. Lei de Moore e Políticas
de Inclusão Digital. Inteligência Empresarial, Rio de Janeiro, v. 14, p. 4-9
NERI, M. C., MELO, L. C. C. de, CORSI, A. P., MONTE, S. dos R. M. Inclusão Digital
no Rio de Janeiro. Inteligência Empresarial (UFRJ). , v.14, p.10 - 16, 2003.
O´REILLY, Tim. What is Web 2.0: Design Patterns and Business Models for the Next
Generation of Software. 2005. Disponível em:
http://oreilly.com/pub/a/web2/archive/what-isweb-20.html?page=1
OCDE Report. Latin American Economic Outlook 2011: How Middle-Class is Latin
America? OCDE Development Centre, 2010.
OECD. A New Economy? The Changing Role of Innovation and Information
Technology in Growth, Paris. 2000.
OECD. The Economic and Social Impact of Electronic Commerce. Preliminary
Findings and Research Agenda, OECD, Paris, 1999.
OECD. Understanding the Digital Divide. Paris, 2000.
OLLEY, G. S. & PAKES, A. The Dynamics of Productivity in the
Telecommunications Equipment Industry. Working Paper nº. 3977, National Bureau
of Economic Research Cambridge (MA), 1992.
PANZAR, J. & WILDMAN, S. S. “Network Competition and the Provision of
Universal Service.” In: Industrial and Corporate Change 4(4): 711-719, 1995.
93
PARKER, P. M. & RÖLLER, Lars-Hendrick. “Collusive Conduct in Duopolies:
Multimarket Contact and Cross-Ownership in the Mobile Telephone Industry.” In:
Rand Journal of Economics 28 (2): 304-322, 1997.
PARTNERSHIP ON MEASURING ICT FOR DEVELOPMENT. Core ICT Indicators
2010. Genebra: InternationalTelecommunication Union, 2010.
PINHO, J . A. G. Sociedade da informação, capitalismo e sociedade civil: reflexões
sobre política, Internet e democracia na realidade brasileira. RAE, vol. 51, nº 1,
jan./fev. de 2011. pp. 22-38.
QUAYNOR, N. Opportunities for Strategies and Policies at the National Level to
Accelerate the Digital Revolution within Developing Economies. Conference Digital
Inclusion Impact and Challenges of the Networked Economy for Developing Countries,
Berlin, 23-24 Janeiro de 2001.
SARTORI, Giovanni. Homo videns: televisão e pós-pensamento. Bauru: Edusc, 2001.
SILVEIRA, S. A. da. Exclusão Digital - A miséria na era da informação. São Paulo,
Editora
STIGLITZ, J. E. Transparency in Government. In: World Bank: The Right to Tell: The
Role of Mass Media in Economic
STIGLITZ, J.; SEN, A. e FITOUSI, J. Report by the Commission on the Measurement
of Economic Performance and Social Progress. September, 2009.
TOFFLER, A. A terceira onda. Rio de Janeiro: Record, 1980.
TRAIN, K. E.; McFADDEN, Daniel L. & BEN-AKIVA, Moshe. “The Demand for
Local Telephone Service: A Fully Discrete Model of Residential Calling Patterns
and Service Choices.” In: Rand Journal of Economics 18(1): 109-123, 1987.
UIT. Child Online Protection: Statistical framework and Indicators. Genebra:
International Telecommunication Union, 2010. Disponível em http://www.itu.int/pub/D-
IND-COP.01-11-2010/en
UNCTAD. Manual for the Production of Statistics on the Information Economy
2009. Nova York: Conferência das Nações Unidas sobre Comércio e Desenvolvimento,
2009. Disponível em http://www.unctad.org/en/docs/sdteecb20072rev1_en.pdf.
UNITED NATIONS. Division for Public Economics and Public Administration / ASPA
(American Society for Public Administration). Benchmarking E-Government: A
Global Perspective Assessing the Progress of the UN Members States, 2002.
US Department of Commerce. National Telecommunications and Information
Administration (NTIA), Falling through the Net. Defining the Digital Divide, A Report
on the Telecommunications and Information Technology Gap in America, Washington,
1999.
94
WAGNER, F. R. Habilidades e inclusão digital: o papel das escolas. In: CGI.br.
Pesquisa sobre o uso das tecnologias da informação e da comunicação no Brasil 2009.
São Paulo: Comitê Gestor da Internet no Brasil, 2010.
WORLD BANK. Public Expenditures for Poverty Alleviation in Northeast Brazil:
Promoting Growth and Improving Services. World Bank LAC Report, 1998.
WORLD ECONOMIC FORUM. The Global Information Technology Report 2002-
2003. 2003.
95
Bibliography on Sewarage, Water and Electricity Infrastructure Sectors
AMPARO, P.P., CALMON, K. M. N. A. Experiência Britânica de Privatização do Setor
de Saneamento. Texto para discussão no 701. IPEA – Instituto de Pesquisa Econômica
Aplicada. Brasília, jan, 2000. Disponível na internet:
www.ipea.gov.br/pub/td/td_2000/td_701.pdf
ANDRADE, T. A., Aspectos Distributivos na Determinação de Preços Públicos. Rio de
Janeiro: IPEA, 199
ANDRADE, T. A , BRANDÃO, Antônio Salazar P., Waldir J.A. e LOBÃO , Salomão
SILVA , L.Q. da. “Saneamento Urbano: A Demanda Residencial por Água”, Pesquisa e
Planejamento Econômico, vol. 25, dezembro de 1995, pp 427-448.
ARAÚJO, R. Regulação da Prestação de Serviços de Saneamento Básico –Abastecimento
de Água e Esgotamento Sanitário. In: IPEA. Infra-estrutura perspectivas de
reorganização (saneamento). Ministério do Planejamento e Orçamento – IPEA, Instituto
de Pesquisa Econômica Aplicada. Brasília, 1999.
AZEVEDO, E. A. Exclusão sanitária em Belo Horizonte – MG: caracterização e
associação com indicadores de saúde. Belo Horizonte: Escola de Engenharia da UFMG,
2003. 145 p. (Dissertação, Mestrado em Saneamento, Meio Ambiente e Recursos
Hídricos).
BARAT, J. O financiamento da infra-estrutura urbana: os impasses, as perspectivas
institucionais, as perspectivas financeiras. In: IPEA. Infra-estrutura: perspectivas de
reorganização (financiamento), Brasília, 1998.
BARROS, R.; CARVALHO, M.; FRANCO, S. “O papel das transferências públicas na
queda recente da desigualdade de renda brasileira”. In: BARROS, R. P. de; FOGUEL,
M.; ULYSSEA, G. (Orgs.). Desigualdade de renda no Brasil: uma análise da queda
recente. Brasília: Ipea, 2007. Cap. 12, v. 2.
BARROS, R.; CARVALHO, M.; FRANCO, S.; “Pobreza Multidimensional no Brasil”.
Texto para Discussão nº 1227, IPEA. Rio de Janeiro, 2006.
BHATTACHARYYA, A., PARKER, E., RAFFIEE, K. An examination of the effects of
ownership on the relative efficiency of public and private water utilities. Land Economics
197 – 209, 1994.
BRISCOE, J. Intervention studies and the definition of dominant transmission routes.
American Journal of Epidemiology, v.120, n.3, p.449-455, 1984a.
BRISCOE, J. Water supply and health in developing countries: selective primary health
care revisited. American Journal of Public Health, v.74, n.9, p. 1009-1013, Sept. 1984b.
BRISCOE J 1985. Evaluating water supply and other health programs: short-run vs long-
run mortality effects. Public Health 99 (3): 142-145.
BRISCOE, J., FEACHEM, R.G., RAHAMAN, M.M. Evaluating health impact; water
supply, sanitation, and hygiene education. Ottawa: International Development Research
Centre, 1986. 80p.
96
BYRNES, P., GROSSKOPF, S., HAYES, K. Efficiency and ownership: further evidence.
Review of Economics and Statistics, 337 – 341, 1986.
CALDERON et al (2004) The Effects of Infrastructure Development on Growth and
Income Distribution, World Bank Policy Research Working Paper No. 3400
CALMON, K. N.; AMPARO, P.P.; MORAIS, M. P.; FERNANDES, M. Saneamento: as
Transformações Estruturais em Curso na Ação Governamental, 1995 - 1998. In:
REZENDE, F.; PAULA T. B. IPEA - Infraestrutura: Perspectivas de Reorganização –
Saneamento. Parte 1. Brasília: Ipea, 1998.
CARIOLA, Eugenio Celedón; ALEGRIA, Maria Angelica. Analisis del proceso de
privatizacion de los sistemas de agua potable y saneamiento urbanos en Chile. REGA, v.
1, n.2, p. 65-85, jul./dez. 2004.
CARMO, C. M.; TÁVORA JÚNIOR, J. L. Avaliação da eficiência técnica das empresas
de saneamento brasileiras utilizando a metodologia DEA. In: Encontro Nacional de
Economia, 31., Porto Seguro, 2003. Anais. Belo Horizonte: ANPEC, 2003 (4 mar. 2004).
CASTAÑEDA, T.; LINDERT, K.; “O desenho e a implementação dos sistemas de
focalização familiar: lições da América Latina e dos Estados Unidos”. World Bank, Séries
sobre Redes de Segurança Social, Abril 2005.
CASTRO, C. E. T. Avaliação da Eficiência Gerencial de Empresas de Água e Esgotos
Brasileiras por Meio da Envoltória de Dados (DEA). Dissertação (Mestrado). Rio de
Janeiro: PUC, Departamento de Engenharia Industrial, 2003.
CHISARI, O., A. ESTACHE e C. Waddams (2003) Price, Access by the poor in Latin
America's utility reform: subsidies and service obligations. In: C. Ugaz and C. Waddams
Price, Editors, Utility Privatization and Regulation: A Fair Deal for Consumers?, Edward
Elgar, Northampton, MA, USA.
CHILE. Superintendencia de Serviços Sanitarios. Evolución del subsidio al pago
delconsumo de agua potable y servicio de alcantarillado de aguas servidas 1990 – 2006.
CORRALES, M. E. La Regulación en Períodos de Transición: el Caso de los Servicios
de Agua Potable y Saniamiento en America Latina. In: BARRIONUEVO FILHO, A.;
LAHERA E. P. (Ed.). Qué Hay de Nuevo en Las Regulaciones? Telecomunicaciones,
Eletricidad y Agua en America Latina. Eudeba/CLAD, 1998.
CORTON, M.L. Benchmarking in the water sector: the case of Peru. Utilities Policy 11,
133e142, 2003
COSSIO, F. A. B. Efeitos das despesas públicas dos estados sobre os indicadores
socioeconômicos estaduais. Instituto de Pesquisa Econômica Aplicada, 2001.
CRAGG, J. G. Some Statistical Models for Limited Dependent Variables with
Application to the Demand for Durable Goods. Econometrica, Vol. 39, No. 5, pp. 829-
844, 1971.
CRAIN,W., ZARDKOOHI, A. A test of the property-rights theory of the firm: water
utilities in the United States. Journal of Law and Economics 395 – 408, 1978.
97
CREMER, F Gasmi, A Grimaud, JJ Laffont (2001), Universal service: An economic
perspective, Annals of Public and Cooperative Economics.
ESTACHE, A., KOUASSI, E. Sector organization, governance and the efficeiency of
African. Utilities Policy Research Working Paper 2890. The World Bank, Washington,
2002.
ESTACHE, A., TRUJILLO, L. Efficiency effects of privatization in Argentina: water and
sanitation services. Water Policy 5 (4), pg. 369 -380, 2003
ESTACHE, A., PERELMAN, S., TRUJILLO, L., 2005. Infrastructure performance and
reform in developing and transition economies: evidence from a survey of productivity
measures. World Bank Policy Research Working Paper 3514, February, 2005
ESTACHE, A., ROSSI, M., 2002. How different is the efficiency of public and private
water companies in Asia? The World Bank Economic Review 16 (1), 139 - 148, 2002
ESTACHE A., GOICOECHEA, Ana, MANACORDA, Marco. Telecommunications
Performance, Reforms and Governance, World Bank Policy Research Working Paper
3822, January 2006.
FARIA, R. C.; SOUZA, G. S.; MOREIRA, T. B. Public Versus Private Water Utilities:
Empirical Evidence for Brazilian Companies. Economics Bulletin, v. 8, n. 2, p. 1-7, 2005.
FEIGENBAUM, S., TEEPLES, R. Public versus private water delivery: a hedonic cost
approach. Review of Economics and Statistics, pg. 672 – 678, 1983.
FERREIRA, C. K. L. O Financiamento da Infra-Estrutura: o Papel do “Project
Financing.” Instituto de Economia do Setor Público. (IESP), jun. 1995.(Texto para
Discussão, n. 25).
FOX, W., HOFFLER, R. Using homothetic composed error frontiers to measure water
utility efficiency. Southern Economic Journal, pg. 461- 477, 1986.
FRYER, R. G.; “Financial Incentives and Student Achievement: Evidence from
Randomized Trials”. Harvard University, EdLabs and NBER, 2010.
FUJIWARA, Thomas. A Privatização Beneficia Os Pobres? Os Efeitos Da
Desestatização Do Saneamento Básico Na Mortalidade Infantil Anais do XXXIII
Encontro Nacional de Economia , 2005.
GALIANI, Sebastian, GERTLER, Paul & SCHARGRODSKY, Ernesto (2005) Water for
Life: the Impact of the Privatization of Water Services on Child Mortality. Journal of
Political Economy, v. 113, n.1.
GARCIA, S., THOMAS, A. The structure of municipal water supply costs: application
to a panel of French local communities. Journal of Productivity Analysis 16, pg. 5 – 29,
2001.
GASMI, F., J.-J. LAFFONT e W. Sharkey, (2000) Competition, universal service and
telecommunications policy in developing countries, Information Economics and Policy
12 (3), 221–248.
98
GLYNN, D. R. et alii. Economic issues in water privatization and regulation. Water
Science and Technology, v. 26, n. 7-8, p. 1921-1928, 1992.
HELLER, L. Saúde e Saneamento. Brasília, OPAS, 90 pp., 1997.
KAKWANI, N. NERI, M. C., SON, H. Linkages Between Pro-Poor Growth, Social
Programs and Labor Market: The Recent Brazilian Experience. World Development.,
v.38, p.881 884,
KAUFMAN, R. R.; SEGURA-UBIERGO, A. Globalização, Política Interna e Gasto
Social na América Latina: Uma Análise de Corte Transversal com Série Temporal,
1973-1997. DADOS – Revista de Ciências Sociais, Rio de Janeiro, Vol. 44, nº 3, pp. 435
a 479, 2001.
KREMER, M.; “Randomizes Evaluations of Educational Programs in Developing
Countries: Some Lessons”.
LAFFONT, J-J. (2005) Regulation and Development, Cambridge University Press
LAFFONT, J.J., TIROLE, J. A Theory of Incentives in Procurement and Regulation. MIT
Press, Cambridge, MA, 1993.
LARSEN, Bjorn; STRUKOVA, Elena. Cost-Benefit Analysis: Water Supply, Sanitation,
and Hygiene and Indoor Air Pollution Interventions in Peru. Paper commissioned for the.
Guatemala CEA, Latin America and Caribbean Region, 2005.
LIBÂNIO, P.A.C. et al. A Dimensão da Qualidade da Água: Avaliação da Relação entre
Indicadores Socias, de Disponibilidade Hídrica, de Saneamento e de Saúde Pública.
Engenharia Sanitária e Ambiental, V. 10, N. 3, p. 219-228, jul./set., 2005.
LINDERT, K.; LINDER, A.; HOBBS, J.; BRIÈRE, B. de La; “The Nuts and Bolts of
Brazil’s Bolsa Família Program: Implementing Conditional Cash Transfers in a
Decentralized Context.” World Bank. Discussion paper n. 0709. 2007
LYNK, E.L. Privatization, joint production and the comparative efficiencies of private
and public ownership: the UK water industry case. Fiscal Studies 14, 98 -116, 1993.
MARTÍNEZ-ESPIÑEIRA, R.. A Box-Cox double-hurdle model of wildlife valuation: the
citizen’s perspective. Department of Economics, St. Francis Xavier University, US, 2004.
MEGGINSON, W.L., SETTER, J.M. From state to market: a survey of empirical studies
on privatization. Journal of Economic Literature XXXIX pg. 321-389, jun, 2001.
MORAIS, M. P. Saneamento: tão perto e tão longe das soluções. IPEA - Revista desafios
do desenvolvimento, no 41, pg. 30 – 36, por Ricardo Wegrzynovski. Brasília, mar, 2008.
NERI, M. C.; Trata Brasil: Saneamento e Saúde. Rio de Janeiro: FGV, CPS/ Trata Brasil,
2007.
NERI, M. C. “Income Policies, Income Distribution, and the Distribution of
Opportunities in Brazil,” in Lael Brainard In: Brazil as an Economic Superpower?
99
Understanding Brazil’s Changing Role in the Global Economy Washington, D.C :
Brookings Institution Press, 2009, p. 221-269.
NERI, M. C. A Nova Geração de Políticas Sociais In: Política de Gestão Pública
Integrada.1ª Rio de Janeiro : Editora FGV, 2008, v.1, p. 72-86.
NERI, M. C.; Trata Brasil: Saneamento, Educação, Trabalho e Turismo. Rio de Janeiro:
FGV, CPS/ Trata Brasil, 2008.
NERI, M. C.; Saneamento, Saúde e o Bolso do Consumidor. Rio de Janeiro: FGV, CPS/
Trata Brasil, 2008.
NERI, M. C., (org) Microcrédito O Mistério Nordestino e o Grameen Brasileiro Perfil e
performance dos clientes do CrediAmigo.1 Rio de Janeiro : FGV, 2008, v.1, p. 260-287.
NERI, M. C. 20 Anos de Constituição: A Hora dos Deveres Sociais In: 20 anos de
Constituição Cidadã: Avaliação e Desafios da Seguridade Social Brasília : Associação
Nacional dos Auditores Ficais da Receita Federal do Brasil (ANFIP), 2008, v.1, p. 249-
254.
NERI, M. C.; Trata Brasil: Saneamento nas Maiores Cidades Brasileiras. Rio de Janeiro:
FGV, CPS/ Trata Brasil, 2009.
NERI, M. C., BUCHMANN, G. From Dakar to Brasilia: monitoring UNESCO’s
education goals. Prospects (Paris). , v.38, p.415 -415, 2009.
NERI, M. C., SOARES, W. L. Estimando o Impacto de renda na Saúde através de
programas de transferência de renda aos idosos de baixa renda no Brasil. Cadernos de
Saúde Pública (FIOCRUZ). , v.23, p.1845 -1856, 2007.
OHIRA, T. H. Fronteira de eficiência em serviços de saneamento no Estado de São
Paulo. Piracicaba, 124 p. Dissertação (Mestrado) – Escola Superior de Agricultura “Luiz
de Queiroz”, Universidade de São Paulo, 2005.
OZUNA J, T; GÓMEZ, I. A. Governance and Regulation: Decentralization in Mexico’s
Water Sector. In: Spiller, P. e Savedoff, W. Spilled. Water: Institutional Commitment in
the Provision of Water Services. Inter-American Development Bank, Latin American
Research Network, Washington- DC, 1999.
PEIXOTO, J. B.. O barulho da água: os municípios e a gestão dos serviços de
saneamento. Água e Vida, São Paulo, 1994.
PEREIRA, C.; MUELLER, B. Comportamento Estratégico em Presidencialismo de
Coalizão: As Relações entre Executivo e Legislativo na Elaboração do Orçamento
Brasileiro. DADOS – Revista de Ciências Sociais, Rio de Janeiro, Vol. 45, nº 2, pp. 265
a 301, 2002.
PEREIRA, Dilma S. P., EMERENCIANO, Emerson Medeiros, BALTAR, Luiz Antonio
de Andrade, Subsídio na Prestação de Serviços de Água e Esgotos. Trabalho apresentado
no I Simpósio de Recursos Hídricos do Centro Oeste Brasilia – 2000.
100
PMSS/SNIS. Programa de Modernização do Setor de Saneamento, Sistema Nacional de
Informações sobre Saneamento -Um extrato do Diagnóstico dos Serviços de Água e
Esgotos. Brasília, Ministério das Cidades, 2003.
PNSB – 2000/IBGE. Pesquisa Nacional de Saneamento Básico 2000. Brasília, IBGE,
2002.
PNUD, IPEA, FJP, IBGE. “Atlas do Desenvolvimento Humano no Brasil”. Rio de
Janeiro: PNUD, 1998.
SAIANI, C. C. S. Restrições à Expansão dos Investimentos em Saneamento Básico no
Brasil: Déficit de Acesso e Desempenho dos Prestadores. (Dissertação de Mestrado) -
Faculdade de Administração, Economia e Contabilidade, Universidade de São Paulo,
Ribeirão Preto, 2007.
SELDEN, T. M. e SONG, D. (1994). Environmental Quality and Development: Is There
a Kuznets Curve for Air Pollution Emissions? Journal of Environmental Economics and
Management, v. 27, n.2, p. 147-162, 1994.
SEROA DA MOTTA, R. Questões regulatórias do setor de saneamento no Brasil. Notas
técnicas do IPEA, Rio de Janeiro, no 5, jan, 2004.
______. (2007). Critérios de Alocação de Recursos no Setor de Saneamento no Brasil. V
Congresso Brasileiro de Regulação, Recife, 06-09 maio de 2007, 2007.
SEROA DA MOTTA, R., AVERBURG, A. Evaluation of the World Bank assistance to
the W&S sector in Brazil. Mimeo, 2002.
SEROA DA MOTTA, R. MOREIRA, A. Efficiency and regulation in the sanitation
sector in Brazil. Utilities Policy 14, 185 -195, mar, 2006.
______. Efficiency and regulation in the sanitation sector in Brazil, trabalho apresentado
na Jornada de Estudos de Regulação do IPEA, Rio de Janeiro, 2004.
______. Eficiência na gestão municipal no Brasil. Texto para discussão n° 1301, IPEA,
Rio de Janeiro, 2007.
SEROA DA MOTTA, Ronaldo; MENDES, Ana Paula. F. Custos de saúde associados à
poluição do ar no Brasil. Rio de Janeiro: Texto de discussão nº 332, IPEA, 1992; p.1-40.
SEROA DA MOTTA, Ronaldo. MENDES, Ana Paula Fernandes, MENDES F. Eduardo,
YOUNG, Carlos E. Firckmann. Perdas e serviços ambientais do recurso água para uso
doméstico. Texto para Discussão nº 258, IPEA, 1992. p.1-40.
SCHNEIDER, A.; RAMIRES, V. R.; “Primeira Infância Melhor: uma inovação em
política pública.” UNESCO, 2007.
TAUCHMANN, H., CLAUSEN, H. Do organizational forms mater? An econometric
analysis of innovativeness in the German wastewater sector. RWI Essen, Discussion
Paper 22, 2004.
101
TORERO, M., CHOWDHURY, S., Bedi, A.S. (2005). Telecommunications
Infrastructure and Economic Growth: A Cross-Country Analysis”. In: TORERO, M.,
Braun J.v. (eds.). Information and Communication Technology for Development and
Poverty Reduction: The Potential of Telecommunications. Baltimore: Johns Hopkins
University Press. p. 125-144.
TRAIN, K. E., Optimal Regulation – The Economic Theory of Natural Monopoly, MIT
Press, 1995.
TRÉMOLET, S., Barbara Evans. Output-Based Aid for Sustainable Sanitation, OBA
Working Paper Series, GPOBA Paper nº 10, September 2010
TUPPER, H.C., RESENDE, M. Efficiency and regulatory issues in the Brazilian water
and sewage sector: an empirical study. Utilities Policy 12, pg. 29- 40, 2004.
TUROLLA, F. A. “Política da senamento básico: avanços recentes e opções futuras de
políticas públicas”. Textos para Discussão do IPEA, Brasília, no 922: p. 1-26, dez, 2002.
TUROLLA, F. A.; OHIRA, T. H. Saneamento básico: experiência internacional e
avaliação de propostas para o Brasil, CNI, Brasília, 2006.
VICKERS, J.. YARROW, G. Privatization: an Economic Analysis. Cambridge - MA:
Massachussetts Institute of Technology, 1988.
WORLD BANK, Brazil - Managing Pollution Problems - The Brown Environmental
Agenda. Volume I e II. Washington D.C. USA. February 27, 1998.
WORLD BANK. “Brasil-Desenvolvimento da Primeira Infância: foco sobre o impacto
das pré-escolas”. Brasília: Banco Mundial/Departamento de Desenvolvimento Humano.
Brasil/Unidade de Gerenciamento do País, Região da América Latina e do Caribe, 2002.
WOOLDRIDGE, J. M.. Introductory Econometrics. Pioneira Thomson Learning, 20 ed,
2002.
YOUNG, M. E. (Ed.); “Early Child Development: from measurement to action”.
Washington, D.C.: The World Bank, 2007. 218 p.
102
Bibliography on Urban Transportation Infrastructure Sector
CALDERON et al, The Effects of Infrastructure Development on Growth and Income
Distribution, World Bank Policy Research Working Paper No. 3400, 2004.
CREMER, F GASMI, A GRIMAUD, JJ LAFFONT, Universal service: An economic
perspective, Annals of Public and Cooperative Economics, 2001.
DOWNS, A., Still Stuck in Traffic: Coping with Peak-Hour Traffic Congestion,
Brookings Institution Press, 2004.
DUMOND, J. M., HIRSCH, B.T., e MACPHERSON, D.A., Wage Differentials Across
Labor Markets and Workers: Does cost of living matter?, 1999.
EHRENBERG, R.G. e SMITH, R.S. (2008). Modern Labor Economics: Theory and
Public. Addison Wesley, 10th edition.
ESTUPIÑÁN, N., GÓMEZ-LOBO, A , MUÑOZ-RASKIN, R. and SEREBRISKY, T.
Affordability and Subsidies in Public Urban Transport: What De We Mean, What Can Be
Done? Policy Research Working Paper #4440. World Bank, Washington, D.C., 2007.
KAKWANI, N., NERI, M. SON, H. Linkages between Pro-Poor Growth, Social
programmes and the Labour Market in Brazil with, World Development, 2010.
LAFFONT, J-J. Regulation and Development, Cambridge University Press, 2005.
LITMAN, T., Evaluating Transportation Equity, Victoria Transport Policy Institute,
Victoria, 1995.
NERI, M Impactos Sociais do Bilhete Único Intermunicipal no Grande Rio, mimeo, Rio
de Janeiro: FGV/CPS, 2010.
SMALL, K., VERHOEF, E., The Economics of Urban Transportation, Routledge,
Abingdon, England, 2007.
SMALL, K.A., Urban transportation policy: A guide and road map, forthcoming in
Robert P. Inman, ed., The Urban Enigma: City Problems, City Prospects, Princeton
University Press, 2008.
TORERO, M., CHOWDHURY, S., BEDI, A.S. Telecommunications Infrastructure and
Economic Growth: A Cross-Country Analysis. In: Torero, M., Braun J.v. (eds.).
Information and Communication Technology for Development and Poverty Reduction:
The Potential of Telecommunications. Baltimore: Johns Hopkins University Press. p.
125-144, 2005.
WORLD BANK., Safe, Clean, and Affordable… Transport for Development — The
World Bank Group’s Transport Business Strategy 2008-2012. Washington, D.C. 2008.
WORLD BANK. Cities on the Move – A World Bank Urban Transport Strategy Review.
Washington, D.C. 2002.
103
APPENDIX: Microdata Sources and Econometric Techniques
This appendix details the different statistical techniques used in the analysis, such as
logistic regression, applied to discrete variables for example in the case of indicators of
access to infrastructure, as well as log-linear income equations. We also detail the
difference in difference estimator and the stepwise methodology applied to these models.
A. Database Description
i. Demographic Census
The sample of the demographic census is a household survey that seeks to interview a
portion of the Brazilian population throughout the national territory (ranging from 25%
in the 70th Census to 10% in the 2000 Census, reaching a variable value in the 2010
Census, nor inversely related to demographic density). This is a survey of occupied
households.
The Census details personal and occupational characteristics of all household members
and has detailed information about the sources of income, access to housing, public
services, trasportation and durable goods, among others. The Census allows analyzing
livin conditions of the population and their determinants at the spatially disaggregated
level. The Census also allows analyzing the long-term trends living conditions of the
population.
ii. National Household Sample Survey (PNAD)
Besides the Demographic Census, there are two main sources of household data at a micro
level that can be used to evaluate at least at an annual frequency the evolution of per capita
income distribution and living conditions in Brazil: PNAD and PNADC. PNAD offers
the possibility of covering different income sources at a national level. In this respect,
PNADC basically covers labor earnings up to now. However, one must have in
perspective that PNAD presents just one picture at one point in every year that the survey
is carried out. Since PNADC is a monthly survey it can provide a better idea of what
happened during the whole year to a less comprehensive set of variables than PNAD. In
sum, PNAD offers a detailed picture once a year of Brazilian social indicators while
PNADC offers a not so detailed but more updated monthly film of the same object.
The PNAD survey is carried annually by the Brazilian Institute for Geography and
Statistics (IBGE) since 1976 (in practice this is the data available), except for the years
104
when the Census takes place. Its sample involves more than 100 thousand families per
year and it has information about several demographic and social-economics
characteristics of the population, including features of households, individuals, families
and workers. It is suitable for objective measures of income and education. Every year it
includes a special supplement about one specific topic. The 2004 and 2006 special
supplement on social programs and education and few supplements on ICTs use will be
central here.
iii. Family Budget Survey (POF)
The first Family Budget Survey (POF) carried out by the IBGE took place in 1987-1988
and has the same geographical coverage as the 1995-1996 survey, which included the
Metropolitan Regions of Belém, Fortaleza, Recife, Salvador, Belo Horizonte, Rio de
Janeiro, São Paulo, Curitiba, Porto Alegre, Brasília and Municipality of Goiânia. In 1996,
it had a sample of 16,060 households, where information was obtained from expenses
incurred during different reference periods (seven, thirty, ninety days or six months),
whose information was collected from October 1995 to September 1996.
The survey was carried out in 2002/03 and 2008/09 with its sample encompassed around
50 thousands households for each wave of the survey. POF’s main objective is to
determine the consumption and expenditures structure of the population. Problems with
paying public services bills and other bills in separate. However, POF also includes
questions about the subjective perceptions of the agents, such as the quality of public
services, such as water service, waste collection and electric energy; perceptions about
related problems such as Dark House, Dark Street, Humidity Problems, Environmental
Problems,; among others.
105
B. Microeconometric Techniques
Multivariate Analysis – Methodology
The bivariate analysis captures the role played by each attribute considered
separately in the demand for insurance. That is, we do not take into account possible
and probable interrelations of the explanatory variables. For example, in the calculation
of insurance by state within the Federation, we don’t consider the fact that Sao Paulo is
a richer place than most states, thus should have greater access to insurance. The
multivariate analysis used further ahead seeks to consider these interrelations through a
regression of the many explanatory variables taken together.
Aiming to provide a better controlled experiment than the bivariate analysis, the
objective is to capture the pattern of partial correlations between the variables, interest
and explanatory. In other words, we have captured the relations between the two
variables, keeping the remaining variables constant. This analysis is very useful to
identify the repressed or potential demand as we compared them, for instance, which are
the chances of a person with more education having higher income, if he/she has the
same characteristics as the comparison group.
i. Logistic regression
The type of regression used in our simple discrete variables multivariate regressions, as well as
to estimate differences-in-differences models. Binomial logistic regression is one method used
to study the determination of dummy variables - those composed of only two options of events,
such as "yes" or "no" . For example:
Let Y be a dummy random variable defined as:
Where each iY has a Bernoulli distribution, which probability distribution function is given by:
y-1y p)-1(pp)|P(y
where y identifies the event that occurred and p is the probability of success of the event.
Since this is a sequence of events with Bernoulli distribution, the sum of the number of
successes or failures in this experiment has binomial distribution of parameters n (number of
observations) and p (probability of success). The binomial distribution probability function is
given by:
y-1y p)-1(py
np)n,|P(y
Logistic transformation can be interpreted as the logarithm of the ratio between the odds of
success versus failure, in which logistic regression gives us an idea of the return of a person to
obtain occupation, given the effect of some explanatory variables that will be introduced later,
in particular vocational education.
The bonding function of this generalized linear model is given by the following equation:
K
0k
ikk
i
ii xβ
p-1
plogη
106
Where the probability pi is given by:
K
0k
ikk
K
0k
ikk
i
xβexp1
xβexp
p
The models used here have the objective of identifying the variables related to
the characteristics of interest (response variable). When performing the model
adjustment, it is desired to find, and to identify, the main factors that best describe the
behavior / variation of the characteristics of interest.
The generalized linear model used here is defined by a probability distribution
for the response variable, a set of independent variables (explanatory factors) that make
up the linear predictor of the model, and a bond function between the mean of the
response variable and the linear predictor.
Odds Ratio:
ii. VARIABLES SELECTION
To select the model we used a Stepwise procedure. The final models were
selected step by step, after grouping the factor levels based on the Wald statistic,
including at each step the interactions that produced the greatest decrease in
Deviance, considering the reason test.
iii. Difference in difference estimator
Example of methodology applied to two different periods
In economics, vast research is done analyzing the so-called experiments or quasi-experiments. To analyze a natural experiment it is necessary to have a control group, that is, a group that was not affected by the change, and a treatment group that was directly affected by the event of interest, both with similar characteristics. In order to study the differences between the two groups, pre and post-event data are needed for both groups. Thus, the sample is divided into four groups: the pre-change control group, the post-change control group, the pre-change treatment group, and the post-change treatment group.
The difference between the differences between the two periods for each of the groups is the difference in difference estimator, represented by the following equation:
g3 = (y2,t – y1,t) – (y2,c – y1,c)
2
2
1
1
p-1
p
p-1
p
107
Where each y represents the mean of the studied variable for each year and group, with the subscript number representing the sample period (1 for before the change and 2 for after the change) and the letter representing the group to which the data belongs (c for the control group and t for the treatment group). g3 is the so-called difference in difference estimator. Once the g3 is obtained, the impact of the natural experiment on the variable to be explained is determined.
In order to study the impacts of local infrastructure policies between two groups, we need data at least two moments in time for both of them. Our sample is thus four fold. The interactive effect between the treatment group dummy (dT=1; dT=0 (control group omitted category)) and the time dummy (d2 =1; d2=0 (initial instant omitted category), which as we will see gives us the difference-in-difference estimator.
Mathematically, we can represent this difference-in-difference estimator (D-D) used from equations in discrete or continuous variables (for example, in the case of logistic regressions or mincerian-type per capita income equations):
Y = g0 + g1*d2 + g2*dT+ (D-D)*d2*dT + other controls
iv. Mincerian Income Equation
The mincerian equation of wage determination is the basis of an enormous literature on
empirical economics. Jacob Mincer's (1974) wage model is the framework used to
estimate returns to education, returns to quality of education, returns to experience, and
so on. Mincer developed an income equation that would be dependent on explanatory
factors associated with schooling and experience, as well as possibly other attributes,
such as gender, for example. It is the basis of education economics in developing
countries and its estimation has already motivated hundreds of studies. It is also used to
analyze the relationship between growth and level of schooling of a society, as well as
effects on inequality. We incorporate variations of this model using infrastructure
variables as possible determinants.
One of the great virtues of the Mincerian equation is to incorporate a single
equation into two distinct economic concepts:
(a) a price equation revealing how much the labor market is willing to pay for
productive attributes affected by infrastructure coverage.
(b) The premium rate of infrastructure,.
REGRESSION MODEL
The typical econometric regression model derived from the Mincerian equation is:
ln w = β0 + β1 educ + β2 hc + β3 hc² + γ′ x + є
where
w is the wage received by the individual,
infra is a vector of categorical variables related to the access of different
infrastructure elements
hc is a vector of quadratic terms for human capital related variables such as years
of schooling, experience (approximated by the age of the individual and household size
108
x is a vector of other observable characteristics of the individual, such as race,
gender, region.. and
є it's a stochastic error
The Coefficient and Attribute Prize
This is a regression model in the log-level format, that is, the dependent variable, the
income is in logarithmic format and the most relevant independent variable here,
infrastructure, is in level format. Therefore, the coefficient β1 measures how much the
coverage of a specific item causes in proportional variation in the wage of the individual.
For example, if sewerage coverage component of vector β1 is estimated at 0.18, this
means that granting sewerage access is related on average with a wage increase of 18%.
This corresponds to the premium of the attribute (or rate of return if the costs were zero).
Mathematically, we have:
Deriving, we find that: ( ∂ ln w / ∂ infra )= β1
On the other hand, by the chain rule, we have:
( ∂ ln w / ∂ infra ) = ( ∂ w / ∂ infra ) ( 1 / w ) = ( ∂ w / ∂ infra ) / w)
Thus, β1=(∂w/∂educ)/w, corresponds to the percentage variation of the wage from a
increase of one year of study..
The coefficient of the mincerian regression with only the constant and a specific variable,
say sewerage coverage, gives the gross or uncontrolled relative premium in terms of
income variation.
The coefficient of a variable of a multivariate mincerian regression (that is, a log-linear
equation with a constant and a series of additional variables) gives us the marginal
controlled relative premium in terms of income variation. Thus, a tentative to isolate the
effect of this variable from the possible correlations with the other variables considered.