FEAR OF CRIME, VIOLENCE AND SCHOOLPERFORMANCE: AN ANALYSIS OF THE

PROBABILITY OF SCHOOL FAILURE IN BRAZIL

Wander Plassa∗ Marina Silva da Cunha†

ResumoO objetivo deste artigo é estudar os efeitos da insegurança e violência sobre a probabilidade de atraso escolar dos adolescentese jovens brasileiros considerando caracterı́sticas geográficas, pessoais e familiares. Para tanto é utilizado dados da PesquisaNacional por Amostra de Domicı́lios (PNAD) de 2009 e o modelo econométrico próbite para obter as probabilidades de atrasoescolar. Além de variáveis relacionadas às caracterı́sticas do aluno e de sua famı́lia, os resultados evidenciam que aqueles queapresentam insegurança em seus próprios bairros ou foram vitimados possuem maiores probabilidades de atraso escolar emque os efeitos da agressão fı́sica seriam maiores sobre mulheres e os mais vulneráveis socioeconomicamente.

Palavras-chave: Insegurança, violência, atraso escolar.

Jel: C01, C25, I21

Área 8: Econometria

AbstractThe purpose of this article is to study the effects of fear of crime and violence on the likelihood of school failure amongadolescents and young Brazilians considering geographical, personal and family characteristics. For that, are used data fromthe National Household Sample Survey - Pesquisa Nacional por Amostra e Domicı́lios (PNAD) - of 2009 and the econometricprobit model to obtain the probability of school failure. In addition to variables related to the characteristics of the studentand his family, the results show that those with fear of crime in their own neighborhoods or were victimized have higherprobabilities of school failure wherein the effects of physical assault would be greater, especially, on women and the mostvulnerable socioeconomically.

Keywords: Fear of Crime, victimization, school failure.

Jel: C01, C25, I21

Area 8: Econometrics

∗Doutorando na Faculdade de Economia, Administração e Contabilidade de Ribeirão Preto da Universidade de São Paulo.E-mail: wanderplassa@outlook.com

†Professora Titular no Departamento de Economia da Universidade Estadual de Maringá. E-mail: mscunha@uem.br

1

mailto:wanderplassa@outlook.commailto:mscunha@uem.br

1 IntroduçãoNowadays one of the biggest concerns of the Brazilian educational system is the inadequate progress of

the individual throughout the school system, due to the number of grade repetition or even the early dropout(Gomes-Neto e Hanushek (1994); Leon e Menezes-Filho (2002)). The Programme for International Stu-dents Assessment - PISA (2012) evaluated 65 countries and found that, for Brazil, despite the improvementin the school grades in the period 2003 to 2012, and the greater inclusion of children in schools, the countrystill has problems with the school failure (the combination of repetition and dropout rates), in which morethan a third of 15-year-old students (36%) repeated a grade at least once the elementary or middle school.Moreover, there was a significant number of students who repeated more than once. Another point notedwas that the proportion of young people who repeated grade in the years 2003 to 2012 remained stable.1

Beyond the question of failure, it is observed the increase in violence and fear in Brazil, where note thatchildren, adolescents and young people are among the most victimized populations (Malta et al. (2010);Delaney-Black et al. (2002)). Homicide rates in the country are already among the highest in the world,being that 19 Brazilians cities compose the rank of the 50 most violent in 2014.2 In the last BrazilianYearbook of Public Safety, for 2013, the violent deaths in Brazil reached more than 53,000 people, with26 homicides per 100,000 inhabitants, which is a record. The homicide rate is another component of thevarious types of crimes (burglary, theft, assault, drug trafficking, etc.) that have been part of the Brazilianreality, especially in big cities.

According to Guimarães (1998), an environment that has been a target of such violence are schoolsand neighborhoods that surround them. Violence and fear in these places have been aggravated due to theincrease of drug trafficking, gang formation and ease to bear arms. This fact is also observed in the studycarried out by UNESCO, entitled “Violências nas Escolas”3 by the researchers Miriam Abramovay, MartaAvancini and Helena de Oliveira to 13 Brazilian capitals more Federal district, in 2001.4 It was verifiedthat is in the school surroundings where the greatest situations of violence occurs. Although the number ofprofessionals and students who indicated the educational institution as more violent than their surroundingsattained a considerable average of 20% of respondents. They also reported being the formation of gangsand drug trafficking within and outside the educational institution as a major problem and responsible forthe climate of fear in school environments.

The report also observed that teenagers or young with fear because of the high violence might have im-paired their school performance, resulting in a probable grade repetition since students who are unsafe mayhave difficulty concentrating or even attend the school properly. Almost half of the respondents indicatedthat their concentration is compromised because of the violence in this environment and an average of 30%of the students in these capitals pointed to the desire to avoid school.

Finally, with more recent information of Brazil, the Pesquisa Nacional de Saúde do Escolar – PENSE(2012) (National Survey of School Health, 2012), conducted by the Instituto Brasileiro de Geografia eEstatı́stica – IBGE (The Brazilian Institute of Geography and Statistics), covers more than 100,000 studentsin the public and private school systems. Of this total, nearly 10% of them indicated that they missed schoolat least once in the last 30 days because they do not feel safe in this environment. Thus, it is possible to notethat students who fear some kinds of violence in schools, or have been victimized are most likely to stay athome for reasons beyond diseases, as shown by Pearson e Toby (1991) and Lab e Whitehead (1992).

Thus, due to this context of violence and fear of crime present in the Brazilian way of life, assesses1 Between 2003 and 2012, the proportion of 15 year-old who had repeated a number in primary school declined, but the

prevalence of grade repetition increased in secondary education. Overall, the percentage of students who had repeated agrade at least once remained stable.

2 Elaborated by the NGO Citizen Council for Public Social Security and Criminal Justice in Mexico 〈http://www.seguridadjusticiaypaz.org.mx/biblioteca/prensa/send/6-prensa/198-las-50-ciudades-mas-violentas-del-mundo-2014〉. Ac-cessed: November 20, 2015.

3 Abramovay, Avancini e Oliveira (2002).4 Goiânia, Cuiabá, Manaus, Belém, Fortaleza, Recife, Maceió, Salvador, Vitória, Rio de Janeiro, São Paulo, Florianópolis and

Porto Alegre.

2

http://www.seguridadjusticiaypaz.org.mx/biblioteca/prensa/send/6-prensa/198-las-50-ciudades-mas-violentas-del-mundo-2014http://www.seguridadjusticiaypaz.org.mx/biblioteca/prensa/send/6-prensa/198-las-50-ciudades-mas-violentas-del-mundo-2014

whether both the violence and fear are factors that contribute to the increased likelihood of school failure.The aim of this paper is to analyze the determinants of school failure between adolescents and young inBrazil, especially the effects of fear of crime and violence. This work seeks to contribute to the nationalliterature on the analysis of school performance in Brazil, addressing a topic that is still little explored,violence and especially fear of crime.

This article is divided into four sections, besides this introduction and conclusion. The first sectionpresents a review of the literature on the effects of violence on school performance. Subsequently, it isexposed the econometric method and the description of the data employed in the paper. Following, the thirdsection provides the descriptive analysis of the school failure behavior and violence. Finally, it is presentedthe results and contributions of the study.

2 THEORETICAL AND EMPIRICAL FOUNDATIONAccording to Grogger (1997), one of the precursors that analyzes the effects of violence and fear on

school performance, the main effect of violence on the students are in their concentration at school, whichwould affect their school performance, generating an effect negative in the acquisition of human capitalfor these young people, in turn, affecting future earnings. Bowen e Bowen (1999) state that both thefear observed in the neighborhood as seen in the school present a negative effect on student performance.The authors also show that children and young people are the most vulnerable to exposure to violence.Furthermore, the works of Grogger (1997) and Bowen e Bowen (1999) analyze the relationship betweenschool performance and school violence, i.e., the violence that occurs within the domains of school andin their neighborhood. On the other hand, there are studies that verify the relationship between schoolperformance and violence in society, known as community violence. The violence considered in this aspectis that observed regardless of location. In this group, includes the works of Delaney-Black et al. (2002),Henrich et al. (2004) as well as the present article.

In addition, studies examining the determinants of fear of crime and consequences both fear and violenceindicate that some groups (women, the poor, the sick, etc.) are more sensitive to their effects. Examplesof authors of these studies are Hale (1996), Skogan e Maxfield (1981), Ferraro (1995), among others. Thehypothesis is that the most vulnerable are less able (or feel less able) to defend themselves against crime.The same hypothesis is raised in this article. It is expected that adolescents and young people in lowersocioeconomic groups or women be more affected by crime in school failure than their counterparts topossess less defense mechanisms.

Grogger (1997), considering the violence in schools and neighborhoods in the United States for the years1984 and 1986, examined how the local violence affects high school graduation and college attendance. Itwas found that, on average, moderate levels of violence reduce the likelihood of high school graduation at5.10% as well as the likelihood of college attendance in 6.90%. Bowen e Bowen (1999), also analyzingdata for the United States, investigated the effects on the proficiency and attendance of students, given theirperception about violence in schools and in their surroundings, in 1996 and 1997. The results obtainedshowed that both fear at school and around it negatively affects the proficiency of the student and schoolattendance, where the effect of violence in the surrounding was more perceived.

Mudege, Zulu e Izugbara (2008) state that children and their parents sometimes find difficult to attendschool because of lack of physical safety guarantees in this environment. This study examined the effectsof fear of crime on the performance of the student and discussed how the perception of safety may affectschool attendance, as well as enrollment of students. The study was conducted for disadvantaged urbanneighborhoods in Nairobi, Kenya, for the year 2004. The results of the study suggest that fear of crime inthese poorer neighborhoods may have negative impact on education.

In another study for the United States, with data for the year 2005, Milam, Furr-Holden e Leaf (2010)found that in addition to the variables that, according to the empirical literature, already have proven effecton school failure of adolescents and youth, the variables related to fear and violence are also relevant.The authors concluded that an increase in violence in the neighborhood was associated with a statistically

3

significant decrease of 4.2% to 8.7% in math and reading scores, while increasing in perceived safety wasassociated with significant increases of 16% to 22% on the score of these subjects.

For Brazil, Severnini (2007), using data for the country in 2003, Teixeira e Kassouf (2011) for the stateof São Paulo in 2007 and Gama e Scorzafave (2013) to the city of São Paulo in 2007, are authors whodiscussed the issue and concluded that there are negative effects of violence on the student’s performancein these locations. According to Severnini (2007) the proficiency of students that attending schools withhigher levels of violence is lower even after the control performed by individual characteristics, as well asteachers and schools. Besides this direct effect, it was found that violence in these schools is related to thehigher turnover of teachers in the school year. Factor that would affect the quality of education in thoseschools.

Teixeira e Kassouf (2011) concluded that violence in schools decreased by 0.54% probability of a stu-dent who attends the third year of high school have a satisfactory performance in mathematics. Gama eScorzafave (2013), analyzing the violence surrounding the school, found that a 10% increase in the murderrate reduces school proficiency at around 0.12 points, both in Portuguese and mathematics subjects. As inSevernini (2007), these results were found even after the use of control variables.

Thus, checking both levels of crime and fear of crime and still evident problem of school failure, thisstudy sought to work on the results found by Grogger (1997). His findings did suggest that interestedresearch in educational production function would do well to focus less on traditional measures such asschool quality and class sizes, and more in less traditional measures that received little attention in the past,for example, fear of crime and violence witnessed by adolescents and youth.

3 Methodological reference

3.1 Discussion on the database and the variables used in this articleThis study aims to consider variables that, based on international and national literature, are able to

capture the effects on student progress in the educational system in Brazil. Beyond the variables traditi-onally established, the objective of this article is to verify if the fear of crime and violence can affect thedevelopment of the student in the educational system.

The database of this article which was used to measure the school failure in Brazil is the PNAD – whoseabbreviation will be used in the tables analyzed - (Pesquisa Nacional por Amostra de Domicı́lio - BrazilianNational Domicile Sample Survey), 2009, which addresses the issue of victimization and fear of crime inBrazil. Prepared as a supplementary nature, the theme victimization was previously carried out only on oneoccasion by PNAD, the year of 1988. More recently, specifically in 2009, this issue was raised. However,it is the first time that this research addresses the issue of public insecurity. PNAD aimed, among otherobjectives, to analyze some behavioral issues associated with victimization event, such as fear of crime andattitudes to prevent violence.

The sample of this article is made up of young people aged 10-23 years that attending school or not,that is, the regular basic education. The dependent variable, school failure, was built based on the paperof Machado (2007). School failure is formed by the following groups: considered a binary variable equalsone for all teenagers or young sample that: a) dropout from school (those who attended school, but do notattend school anymore and not completed high school); b) repeated (attended school, but do not have theyears of study consistent with age), and; c) not in school and never attended before; and zero, otherwise.

Table 1 lists the variables used in this study and their expected signals, based on the literature. Regardingthe variable color, those who are white or yellow formed a group, white, and other ethnicities (black andmulatto) formed another group, non-white. Concerning the variables work and pregnancy in the schoolperiod, the first assumes value one for individuals who worked or work in the age group 10-17 years andzero otherwise. The second, addresses only women, assumes value one for those who became pregnantbetween 10 and 17 years and zero otherwise.

4

Tabl

e1

–E

xpec

ted

sign

als

tofo

rthe

vari

able

sus

edin

the

artic

le

Var

iabl

esC

ode

Exp

ecte

dSi

gnal

sE

xpla

natio

nL

itera

ture

Dep

ende

ntva

riab

leSc

hool

Failu

reM

acha

do(2

007)

Geo

grap

hica

lcha

ract

eris

tics

Bra

zilia

nre

gion

s1

-oth

erre

gion

s,0

-Nor

thN

egat

ive

Res

iden

tsof

nort

hern

Bra

zila

rem

ore

likel

yto

scho

olfa

ilure

Mac

hado

(200

7);

Tria

cae

Teja

da(2

014)

Pers

onal

char

acte

rist

ics

Col

or1

-non

-whi

te,

0-w

hite

Posi

tive

Be

non-

whi

tein

crea

ses

the

likel

ihoo

dof

scho

olfa

ilure

Alb

erna

z,Fe

rrei

rae

Fran

co(2

002)

Sex

1-m

an,

0-w

oman

Posi

tive

Bei

nga

man

incr

ease

sth

epr

obab

ility

ofsc

hool

failu

reM

acha

doe

Gon

zaga

(200

7);

Aqu

ino

ePa

zello

(201

1)W

ork

orw

orke

din

the

scho

olpe

riod

1-y

es,

0-n

ãoPo

sitiv

eW

orki

ngin

crea

ses

the

prob

abili

tyof

scho

olfa

ilure

Mac

hado

(200

7);

Psac

haro

poul

os(1

997)

Preg

nanc

ydu

ring

scho

olpe

riod

1-y

es,

0-n

ãoPo

sitiv

ePr

egna

ncy

duri

ngsc

hool

peri

odin

crea

ses

the

likel

ihoo

dof

scho

olfa

ilure

Mud

ege,

Zul

ue

Izug

bara

(200

8)

Fam

ilych

arac

teri

stic

s

Inco

me

brac

ket

1-h

ighe

rinc

ome,

0-l

ower

inco

me

Neg

ativ

eH

ave

low

erpe

rper

son

hous

ehol

din

com

ein

crea

ses

the

prob

abili

tyof

scho

olfa

ilure

Alb

erna

z,Fe

rrei

rae

Fran

co(2

002)

;Po

ntili

eK

asso

uf(2

007)

Mot

her’

sed

ucat

ion

1-h

igh

educ

atio

n,0

-low

educ

atio

nN

egat

ive

Hav

em

othe

rwith

low

educ

atio

nin

crea

ses

the

likel

ihoo

dof

scho

olfa

ilure

Hon

da(2

007)

;C

urri

ee

Mor

etti

(200

3);

Aqu

ino

ePa

zello

(201

1)

Fath

erle

ss1

-yes

,0

-não

Posi

tive

Liv

ing

ina

sing

le-p

aren

tfam

ilyin

crea

ses

the

prob

abili

tyof

scho

olfa

ilure

Tria

cae

Teja

da(2

014)

Num

bero

fSib

lings

Qua

ntita

tive

Posi

tive

Hig

hern

umbe

rofs

iblin

gsin

crea

ses

the

likel

ihoo

dof

scho

olfa

ilure

Hon

da(2

007)

Fear

ofC

rim

eD

omic

ile1

-yes

,0

-não

Posi

tive

Hav

ing

fear

ofcr

ime

incr

ease

sth

elik

elih

ood

ofsc

hool

failu

re

Gro

gger

(199

7);

Bow

ene

Bow

en(1

999)

;M

udeg

e,Z

ulu

eIz

ugba

ra(2

008)

Nei

ghbo

rhoo

dM

unic

ipal

ityFe

arof

Cri

me

Atte

mpt

edR

obbe

ry/T

heft

1-y

es,

0-n

ãoPo

sitiv

eH

ave

been

vict

imof

crim

esin

crea

ses

the

likel

ihoo

dof

scho

olfa

ilure

Mila

m,F

urr-

Hol

den

eL

eaf(

2010

);G

rogg

er(1

997)

;A

bram

ovay

,Ava

ncin

ieO

livei

ra(2

002)

The

ftR

obbe

ryPh

ysic

alas

saul

tM

ultip

leC

rim

es*

Not

e:(*

)Thi

sca

tego

ryen

com

pass

estw

oor

mor

edi

ffer

entt

ypes

ofcr

imes

.So

urce

:Ow

nel

abor

atio

n.

5

Some divisions were made with respect to the income bracket and mother’s education. Divided thehousehold monthly income per person in seven different groups: income1 for those without income up to1/4 of the minimum wage; income2 for individuals with household income per person higher than 1/4 to1/2 minimum wage; the next bracket, income3, groups the incomes exceeding half the minimum wage upto 1 minimum wage; in income4 are those with household income per person of 1 to 2 minimum wages; inincome5 group are the incomes higher than 2 minimum wages to 3; in income6 rents ranging from 3 to 5times the minimum wage minimum wages; and, finally, income7 are those who reported income above 5minimum wages.5

Moreover, mother’s education information were divided into four groups: i) mothers with no educationand incomplete primary education (under four years of study) – mother’s education1; ii) mothers with com-plete primary and incomplete secondary education (four years full studies to less than 11 years of schooling)- mother’s education2; iii) mothers with high school and incomplete college (11 years of schooling comple-ted to less than 15 years of study) - mother’s education3 and, lastly; iv) mothers with college graduate (15years or more of study) - mother’s education4.6

As household infrastructure, both mother’s education as income are strongly associated with wealth.Thus, house infrastructure was used as an instrument for mother’s education given its endogenous cha-racteristic. Five variables (Possession of movable property, residence condition, locomotion, sewer/trash/water and raw material of the house) have been considered, as described in the appendix of this article,obtained directly from the PNAD of 2009. Honda (2007), point out that the problem of endogeneity in thevariable “mother’s education” emerges as the unobserved variable defined as ability influences the mother’seducation process and the level of school failure.

The econometric model used in this article is the Probit. Model used when the dependent variable isbinary, where y has a value of one for individuals in school failure and zero otherwise. 7 Therefore, this worktried to find the probability of an individual being in school failure given their observable characteristics,described in Table 1.

Prob(Y = 1 | x) = Φ(x′β) (1)

Where Φ is the cumulative standard normal distribution function and x is a row vector of explanatoryvariables (that here defined as geographical, personal and family characteristics, as well as, fear of crimeand victimization) and a constant, and β is the column vector of coefficients. The marginal effect, alsoutilized in this paper is presented as Cameron e Trivedi (2005) as follows,

MarginalEffects =∂pi∂xij

= φ(x′β)βj (2)

Where pi = Φ(x′β). The value of βj gives the proportionate effect on the probability that yi = 1 (i.e.,

the individual is in school failure) as xij changes.

4 ANALYSIS DESCRIPTIVEFrom the description of the data it is possible to evaluate the Brazilian scene as the characteristics of

adolescents and young Brazilians and their families in relation to school failure and its determinants. The5 The minimum wage in the period covered by the survey was R$ 465.00 (about U$ 233.00) Source: 〈http://portal.mte.gov.br/

sal min/〉. Accessed: December 19, 2015.6 Information about the respondents’ mothers are not given directly by the PNAD of 2009. So, to obtain information about edu-

cation, household monthly income per person and number of children, the creation of a key with the following compositionvariables was necessary: V0102 (control number), V0103 (serial number) and V0301 (order number) given by PNAD.

7 For further details about binary choice models see Cameron e Trivedi (2005).

6

http://portal.mte.gov.br/sal_min/http://portal.mte.gov.br/sal_min/

sample studied covers 95,700 individuals who, after using weights available by the PNAD, represents aEstimated population of 45,848,470 people.8

Figure 1 presents the division of the estimated population after made the cuts necessary for the purposeof this work. As can be seen, school failure group is composed of 18,219,954 people (39.74% of total),consisting of 10,289,819 (22.44%) people considered repeaters, 7,620,587 (16.62%) people who were outof school (dropped out) and 309,548 (0.68%) who had never attended school. Thus, there is an indicatorthat refers the teenager’s schooling process.

Figure 1 – Division of estimated population, BrazilSource: PNAD data. Own Elaboration.

In figure 2 is presented by age 10 to 23 years, the behavior of each variable that forms the variableschool failure. Initially, for children of 10 years old, school failure is made up, largely, by people whorepeated a grade. Of the total number of students in failure (20.0%), 19.10% of them repeated the grade,0.80% dropped out of school and 0.50% never attended the Brazilian basic education. The percentage ofpeople in school failure increases with age ranges from 20.40% with 10 years old for 52.65% of those ages19, its maximum value. This means that in Brazil for the year 2009, when considering only individuals 19years, more than half of them would be in school failure. From the 19 years, the percentage of people inschool failure reduces to 42.30% in the age bracket of 23 years.

The repetition rate, primarily responsible for school failure in the early age groups, rises to reach itsmaximum at 14, with 33.35%. From that point is replaced by a downward trend, reaching only 5.92% at theage of 23 years. Part of this decline in the percentage of grade repeat over the years is the question raised bythe authors Cairns, Cairns e Neckerman (1989), Grissom e Shepard (1989), Leon e Menezes-Filho (2002),in which students that are in status “repetition” are more likely to drop out of school. Another explanationis that these people may have concluded high school, which means leaving the group in school failure.

The dropout, from the age of 14, is bigger and has great influence on school failure. The percentageof people dropped out from school exceeds the percentages of individuals in the status of repetition at age18, reaching its maximum at 23 years with 34.96% of total considered school evaded, i.e., they left school8 All calculations presented in this article were conducted considering the estimated population based on the weights provided

by PNAD.

7

before the end of basic education. The variable “never attended school” has slight positive trend over theyears, but its maximum value is only 1.42% at the age of 23 years.

Figure 2 – Failure School by category and age, BrazilSource: PNAD data. Own Elaboration.

It is possible to note that there are three different patterns with respect to the Brazilian school failure.The first pattern, found in the range of 10 to 15 years, the most of school failure is composed of graderepetition. The second pattern observed in the range of 15 to 18, in what is perceived reversal in thecomposition of school failure, where the school dropout becomes the most responsible for the failure. Inturn, the last pattern, 18 to 23, school failure has downward trend, influenced by the end of high school,where there is sharp drop in the percentage of students with repetition and slight increase in the percentageof individuals dropped from school.

In short, the results suggest that in Brazil access to school covers almost the entire school-age popula-tion, however, a considerable part of this population does not progress continuously in the school system,either on account of school retention, on younger groups, either on account of school dropout on older.These results corroborate the Gomes-Neto e Hanushek (1994) and Leon e Menezes-Filho (2002) works.

The table 2 shows the percentages of the population of adolescents and young people that are in schoolfailure in Brazil and also were victimized or not feel safe. The analysis is done by a household incomerange per person where income1 are the poorest and the income7 the richest, according to the methodology.Regarding the types of crimes covered in this article, it is clear that, those who have been victimized and arein income1, more than 50% of individuals were classified in school failure. Physical assault is the kind ofcrime that affects more students in failure in Brazil. The percentage of individuals in the range of income1that are school failure and that reported physical assault reaches 67.92%. In general, as the income willbe increasing the percentage of individuals in failure and that have suffered some kind of crime reducessharply.

To fear of crime in which are analyzed fear on three levels, domicile, neighborhood and municipality, asimilar pattern of victimization was found. In the three levels of fear, about 60% of the population that wasin the income bracket income1 were in school failure, reaching 61.37% for fear of crime indicated in theneighborhood. In the higher income bracket, the percentage falls sharply. For example, only 4.96% of thepopulation that was in the range of income7 and presented fear of crime in the neighborhood was in schoolfailure.

8

Table 2 – Percentage of people in school failure by household income range,according to victimization and fear of crime, Brazil

Description Income bracket1 2 3 4 5 6 7Victimization

Attempted Robbery/Theft (%) 58.05 45.57 36.47 25.42 11.79 8.78 5.93Theft (%) 59.86 50.20 38.20 21.84 15.44 14.92 1.33

Robbery (%) 56.84 54.59 41.21 30.53 15.70 8.16 1.09Physical assault (%) 67.92 71.66 54.29 36.60 13.93 14.82 3.10Multiple Crimes (%) 63.83 53.35 39.99 25.73 19.37 7.94 3.83

Fear of CrimeDomicile (%) 60.64 52.93 38.87 24.47 16.01 7.97 6.13

Neighborhood (%) 61.37 51.09 38.04 23.21 13.91 7.41 4.96Municipality (%) 59.97 49.85 37.28 23.05 13.66 7.86 5.06

Source: PNAD data. Own Elaboration.

It is possible to note that both school failure and victimization/fear of crime affecting a larger numberadolescents and young people in lower income bands in Brazil. This result may indicate that there is arelationship between victimization/fear of crime with school failure, at least in the lower income ranges asnoted by Mudege, Zulu e Izugbara (2008) specifically to fear of crime.

5 RESULTS AND DISCUSSIONFrom the use of the statistical technique of regression probit, it is possible to understand what explana-

tory factors influence the probability of a person is or is not in school failure and if this influence is positiveor negative. Thus, this technique is applied whose results are shown in Table 3. The analysis is made for twospecifications, the first take account variables relating to geographical characteristics, personal characteris-tics and family characteristics. The second, in addition to the variables mentioned in the first specification,adds to analyze the issue of fear of crime and victimization.

Before analyzing the results it was found, from the endogeneity test, that the mother’s education is en-dogenous then the employment of instruments was needed.9 Thus, combined with the analysis of the twospecifications it was also performed the endogeneity correction. The estimated coefficients were mostly sta-tistically significant, with a significance level of 1% and the adjustment values are in line with expectationsin micro analysis.

The first group that is composed of variables relating to geographical characteristics, presents results,mostly expected. For Brazilian regions, it was observed that residents of northern Brazil (reference region)are more likely to be in failure when compared to residents of other localities. The marginal effects indicatethat residents of other Brazilian regions have, on average, about 10% lower chance of failure when comparedto the North. Similar results were found for Brazil in the work of Triaca e Tejada (2014) and Machado(2007).

In the group personal characteristics it is possible to observe attributes such as color, sex, work orworked in the school period and pregnancy in school period. Note that those who are considered non-whiteand man are generally with more likely to be in school failure compared with their counterparts, 0.5%and 14.1%, respectively. The two other aspects that compose the group of personal characteristics presentvery interesting results. For example, those who responded have already worked or be working in theschool period have higher chances of being in school failure. For the variable that considers only women,individuals who became pregnant in the age group 10-17 years are more likely to school failure, about 31%.Psacharopoulos (1997) points out that the number of repetitions, common phenomenon in Latin America,9 Results presented in Appendix of this article in A2 and A3 tables.

9

is closely linked to child labor. Early pregnancy is one of the reasons found in the paper conducted byMudege, Zulu e Izugbara (2008) for school dropout among girls, for example.

The third group of variables analyzed, family characteristics, addresses variables related to householdmonthly income per person, mother’s education, single parent families (fatherless) and, finally, number ofsiblings. The household monthly income per person appears with a very small coefficient, showing a negli-gible marginal effect value. In the mother’s education, however, each year of maternal education reducesthe likelihood of failure in about 6.7%, which corroborates the literature which finds a great relevance inmaternal education on school performance (CURRIE; MORETTI, 2003; AQUINO; PAZELLO, 2011).

Table 3 – Results of the estimation of IV probit models, Brazil

Independent variables (1) (2)Coefficient Robust SE Marginal effect Coefficient Robust SE Marginal effectConstant 0.9447** 0.0016 - 0.9514** 0.0016 -Geographical characteristicsNortheast -0.1902** 0.0008 -0.0682 -0.1920** 0.0008 -0.0689South -0.1902** 0.0010 -0.1227 -0.3587** 0.0010 -0.1229Southeast -0.2999** 0.0008 -0.1227 -0.3587** 0.0008 -0.1076Midwest -0.2391** 0.0011 -0.0831 -0.2386** 0.0011 -0.0830Personal characteristicsNon-White 0.0161 0.0005 0.0059 0.0146** 0.0005 0.0053Man 0.3914** 0.0005 0.1417 0.3888** 0.0005 0.1408Worked in School Period 0.0437** 0.0006 0.0160 0.0408** 0.0006 0.0150Pregnancy in School Period 0.8202** 0.0023 0.3182 0.8129** 0.0023 0.3155Family characteristicsIncome 0.0000** 0.0000 0.0000 0.0000** 0.0000 0.0000Mother’s education*** -0.1838** 0.0001 -0.0672 -0.1848** 0.0001 -0.0675Fatherless 0.2148** 0.0005 0.0802 0.2121** 0.0005 0.0792Number of Siblings 0.0231** 0.0002 0.0084 0.2121** 0.0002 0.0084Fear of CrimeDomicile - - - -0.0489** 0.0008 -0.0178Neighborhood - - - 0.0438** 0.0008 0.0161Municipality - - - -0.0148** 0.0006 -0.0054VictimizationAttempted Robbery/Theft - - - 0.0185** 0.0022 0.0068Theft - - - 0.0518** 0.0018 0.0191Robbery - - - 0.0175** 0.0020 0.0064Physical assault - - - 0.2478** 0.0018 0.0944Multiple Crimes - - - 0.0733** 0.0014 0.0272Estimated population 34,933,686 34,933,686Wald chi2(12) 9388433.72 9484273.88Prob > chi2 0.000 0.000Correctly classified 70.01% 69.98%

Note: control variables omitted from the table; p-value

Although it is expected that the fear of crime in the three levels positively affect the likelihood of schoolfailure, only the fear felt at neighborhood showed this characteristic. Grogger (1997) points out that thisfear indicated in the neighborhood can indicate fear in schools and can inhibit the adolescent or young togo to school, or even have a good academic performance. In addition, other authors (Milam, Furr-Holden eLeaf (2010); Bowen e Bowen (1999)) found significant positive effects of violence and fear of crime in theneighborhood on poor academic performance.

Finally, the group victimization indicates that those who were victims of the crimes listed in this articlegenerally have a higher probability of school failure. Physical assault was the crime that had the highestmarginal effect, 9.4%, which was expected since this crime is one that involves physical contact and therelative severity (SKOGAN; MAXFIELD, 1981).

Table 4 reports the average of the probabilities that an individual has to repetition, dropout and schoolfailure given some selected attributes of personal and family characteristics. With respect to the characteris-tics considered, individuals with the following attributes formed the group that had the highest probability ofschool failure: man, non-white, low income and low mother’s education. People with these attributes havean average probability of 56.52% to repeat, 18.57% to dropout the school and 79.05% of school failure.

Table 4 – Results of Marginal Effects for selected variables of personal characteristicsand family characteristics to Brazil

Personal characteristics Family Characteristics Average Probability ofpresenting school failureSex Color Income Mother’s Education Repeat Dropouts School FailureMan Non-White Low Low 56.52 18.57 79.05Man White Low Low 52.78 17.23 74.71Man Non-White Moderate Low 28.99 32.26 66.69

Woman Non-White Low Low 48.41 12.20 65.10Woman White Low Low 41.34 14.24 59.88

Man White Moderate Low 22.07 27.70 55.54Man Non-White Low Moderate 37.60 5.46 46.88

Woman Non-White Moderate Low 27.48 14.70 45.91Man White Low Moderate 24.24 7.08 38.56

Woman White Moderate Low 19.20 11.60 33.84Man Non-White Moderate Moderate 20.29 6.56 28.30

Woman Non-White Low Moderate 21.37 1.60 26.68Woman White Low Moderate 18.01 3.75 23.06

Man White Moderate Moderate 16.98 4.57 23.01Woman Non-White Moderate Moderate 11.26 2.70 14.36Woman White Moderate Moderate 9.44 2.13 13.25

Average 20.38 8.41 32.26

Note: (*) Low income were defined as those classified in income1 and moderate income individualsclassified as income3 group. (**) Low mother’s education are those mothers classified in Mother’sEducation1, moderate mother’s education are mothers classified Mother’s Education3.

Source: PNAD data. Own Elaboration.

The group formed by women, white, moderate income and moderate mother’s education is the one withthe lowest average probability of school failure and repetition. The average probability of grade repetitionis 9.44% and the school failure is 13.25%. The average probability of dropout was not the lowest observed,however, reached the value of 2.13%. Slightly higher than indicated by the group with these attributes:woman, non-white, low income and moderate mother’s education, of 1.60%.

The total Average Probability presented by the estimated population was 20.38% for grade repetition,8.41% for dropout and 32.26% for school failure. An interesting point is that moderate income integratedthe group with the highest average probability of dropout from school. This result may indicate that theseindividuals may be evading school to join the labor market and thus get a higher income for them andtheir families. It is also noticed that the variable mother’s education seem to play an important role in the

11

average probability of the Repeat, Dropouts and School Failure. Individuals who have moderate mother’seducation appears with lower average probability of school failure comparing with other groups with similarcharacteristics, but with inferior mother’s education.

Seeking investigate the differences observed for school failure by sex in Brazil in Table 5 is analyzedthree specifications, the age group 10-23 years, the age group of 10-15 years (adolescents), which corres-ponds to the first pattern of figure 2, and the range of 15-18 years (young) for the second pattern. Startingthe analysis by geographical characteristics, it is clear that for Brazilian regions, for both men and women,to reside in the North increases the likelihood of school failure regardless of age. However, the differencebetween probability of school failure among residents of other regions and the reference region, Northernregion, seems higher among women. This means to live in a less developed region appears to affect womenmore than men when it comes to school performance.

Regarding personal characteristics, paying attention to the men in non-white attribute, the sign of thisvariable is positive for the 10-23 and 15-18 ages, i.e., be non-white increases the likelihood of school failure.However, for children (10-15 years), the value is strictly close to zero, which indicates that there is not aconsiderable difference between the probability of school failure between whites and non-whites in this agegroup. For women, for the three specifications, be non-white increases the chance of school failure.

The variable that analyzes the work in the school period had divergent effects when comparing the sexes.For men, except for the age group of 15-18 years, the effect of the work is positive about school failure, asexpected. Nonetheless, for women the effect is negative, which indicates that the work in the school perioddecreases the likelihood of school failure among adolescents and young women. Boys are more likely toundertake activities in and in agriculture and industry (hazardous works) while girls outnumber boys inservices. The boys then are more affected by their activities what the results indicate that child labourfurther increase the likelihood of school failure among boys. With respect to early pregnancy, it is clear thatyounger groups are most affected by pregnancy. Moore e Waite (1977) said that girls who became pregnantwhile still quite young themselves tend to complete fewers years of formal schooling than do those whodelay entry into motherhood.

With respect to family characteristics, in the mother’s education, it is possible to see that, regardlessof the individuals sex, increased in mother’s education reduces the likelihood of school failure. For singleparents and number of children, the results suggests that for both men and women, living in these kinds offamily increase the likelihood of school failure. Futhermore, realize that the effects of a family without thefather’s presence seems to be a very important factor in explaining increases the probability of school delay,regardless of gender and age range.

Focusing on groups of interest in this article, fear of crime and victimization, it is possible to note thatwomen are more affected by fear of crime than men. The first set of findings suggests that, in general,fear is statistically significant and positive in to affect the likelihood of school failure among women. Bycontrast, on men only the effect of fear of crime in the neighborhoods for the three age groups was positiveto affect school failure. Moreover, the effects are greater on the groups of younger ages.

Regarding victimization, although some exceptions, the signs are, for the most, positive and statisticallysignificant. This result may reflect that victimization increases the likelihood of school failure for bothsexes in the age groups under consideration. The main crime to affect adolescents and young people seemto be physical assault. This time the older individuals are more affected, but women continue to be the mostsensitive and therefore more affected by this crime.

12

Tabl

e5

–R

esul

tsof

the

estim

atio

nsof

the

spec

ifica

tions

ofsc

hool

failu

reby

sex

and

age,

Bra

zil

Inde

pend

entv

aria

bles

10-2

3ye

ars

10-1

5ye

ars

15-1

8ye

ars

Men

Wom

enM

enW

omen

Men

Wom

enC

oef.

SEC

oef.

SEC

oef.

SEC

oef.

SEC

oef.

SEC

oef.

SEC

onst

ant

1.34

6**

0.00

20.

920*

*0.

002

1.17

6**

0.00

30.

691*

*0.

004

1,61

7**

0.00

31,

127*

*0.

004

Geo

grap

hica

lcha

ract

eris

tics

Nor

thea

st-0

.163

**0.

001

-0.2

23**

0.00

1-0

.169

**0.

002

-0.2

51**

0.00

2-0

.187

**0.

002

-0.2

04**

0.00

2So

uth

-0.3

30**

0.00

1-0

.391

**0.

002

-0.4

98**

0.00

2-0

.501

**0.

002

-0.3

01**

0.00

2-0

.310

**0.

003

Sout

heas

t-0

.279

**0.

001

-0.3

16**

0.00

1-0

.298

**0.

002

-0.3

02**

0.00

2-0

.317

**0.

002

-0.3

41**

0.00

2M

idw

est

-0.2

28**

0.00

2-0

.249

**0.

002

-0.3

12**

0.00

2-0

.289

**0.

002

-0.2

33**

0.00

3-0

.246

**0.

003

Pers

onal

char

acte

rist

ics

Non

-Whi

te0.

020*

*0.

001

0.00

9**

0.00

1-0

.001

**0.

001

0.01

4**

0.00

10.

054*

*0.

001

0.00

1**

0.00

1W

orke

din

Scho

olPe

riod

0.11

2**

0.00

1-0

.115

**0.

001

0.02

7**

0.00

1-0

.113

**0.

002

-0.0

17**

0.00

1-0

.194

**0.

002

Preg

nanc

yin

Scho

olPe

riod

--

0.83

4**

0.00

2-

-1.

073*

*0.

009

--

0.77

1**

0.00

3Fa

mily

char

acte

rist

ics

Inco

me

0.00

0**

0.00

00.

000*

*0.

000

0.00

0**

0.00

00.

000*

*0.

000

0.00

0**

0.00

00.

000*

*0.

000

Mot

her’

sed

ucat

ion

-0.1

87**

0.00

0-0

.180

**0.

000

-0.1

87**

0.00

0-0

.171

**0.

000

-0.1

94**

0.00

0-0

.189

**0.

000

Fath

erle

ss0.

207*

*0.

001

0.22

0**

0.00

10.

220*

*0.

001

0.22

5**

0.00

10.

229*

*0.

001

0.20

7**

0.00

1N

umbe

rofS

iblin

gs0.

016*

*0.

000

0.22

0**

0.00

00.

026*

*0.

000

0.04

3**

0.00

00.

018*

*0.

000

0.04

4**

0.00

0Fe

arof

crim

eD

omic

ile-0

.033

**0.

001

-0.0

63**

0.00

1-0

.048

**0.

002

-0.1

20**

0.00

2-0

.021

**0.

002

0.03

3**

0.00

2N

eigh

borh

ood

0.03

3**

0.00

10.

018*

**0.

001

0.05

9**

0.00

20.

086*

*0.

002

0.04

2**

0.00

2-0

.055

**0.

002

Mun

icip

ality

-0.0

33**

0.00

10.

027*

*0.

001

-0.0

08**

0.00

10.

009*

*0.

001

-0.0

11**

0.00

20.

013*

*0.

002

Vict

imiz

atio

nA

ttem

pted

Rob

bery

/The

ft0.

023*

*0.

003

0.01

0**

0.00

3-0

.125

**0.

006

0.02

9**

0.00

6-0

.004

0.00

5-0

.057

**0.

005

The

ft0.

030*

*0.

002

0.08

4**

0.00

30.

006

0.00

5-0

.035

**0.

006

0.00

40.

004

0.11

5**

0.00

4R

obbe

ry0.

004*

*0.

003

0.02

5**

0.00

30.

032*

*0.

004

0.11

3**

0.00

5-0

.012

*0.

004

0.01

6*0.

005

Phys

ical

assa

ult

0.18

5**

0.00

20.

363*

*0.

003

0.06

2**

0.00

30.

238*

*0.

004

0.22

1**

0.00

40.

347*

*0.

005

Mul

tiple

Cri

mes

0.10

1**

0.00

20.

035*

*0,

002

0.03

6**

0.00

30.

008

0.00

50.

049*

*0.

003

-0.1

18**

0.00

4E

stim

ated

popu

latio

n18

,443

,000

16,4

90,6

868,

985,

781

8,50

9,24

05,

591,

204

5,03

9,82

9W

ald

chi2

(12)

5432

008.

1735

4341

0.58

2171

861.

4813

1440

5.85

1966

512.

2814

5142

9.09

Log

pseu

dolik

elih

ood

-598

4349

6-5

2544

997

-286

5356

6-2

6568

756

-181

0182

8-1

6188

488

Prob

>ch

i20,

000

0,00

00,

000

0,00

00,

000

0,00

0C

orre

ctly

clas

sifie

d67

.14%

73.3

3%69

.35%

75.6

7%67

.12%

70.6

4%

Not

e:co

ntro

lvar

iabl

esom

itted

from

the

tabl

e;p-

valu

e<

0.05

:(*)

,p<

0.01

:(**

);So

urce

:PN

AD

data

.Ow

nE

labo

ratio

n

13

Figure 3, using the marginal effects, show the association between victimization/fear of crime andschool failure for different levels of mother’s education (in years) of the estimated population and diffe-rentiating by sex.10 Regarding parts 3b, 3d and 3f, referring to the fear of crime, you can see that themarginal effect of fear in the neighborhood on the likelihood of school failure is positive and more intensein the lower education. For women, in addition to the fear of crime in the neighborhood, the fear shown atthe municipal level also has positive marginal effect, at least for those with lower maternal education. Othertypes of fear, however, had a different effect than expected.

(a) Victimization (b) Fear of Crime

(c) Victimization in men (d) Fear of crime in men

(e) Victimization in women (f) Fear of crime in womenFigure 3 – Marginal effects of Fear of Crime and Victimization on School Failure

Source: PNAD data. Own Elaboration.

In parts 3a, 3c and 3e is possible to note the relationship of victimization while mother’s educationchanges. It is noticed that, initially for the estimated population total and lower mother’s education, the10 For the variable mother’s education were used instruments for endogeneity correction. These instruments indicate family

wealth.

14

types of crime are positively related to the school failure in that physical assault has a far superior effectcompared to other types of crimes. Furthermore, women are the group more affected by this type of crime.This was expected inasmuch as physical assault is an aggravated physical attack, a crime against a personand can have connection with sexual attacks, for example. Which can produce serious consequences on thefemale victim, as said Hale (1996) and Lipschitz et al. (2000).11

The current findings suggest that people in lower mother’s education are most affected by school failure.In addition to variables in personal and family characteristics, fear of crime and victimization, for instance,were positive on school failure in lower mother’s education, but not, in general, for the highest ones. Thisresult was expected, once individuals with low mothers education are generally poorer and, as said Hale(1996), people with less purchasing power are less able to protect themselves and to avoid situations whichmight produce fear, such as move out from violent schools or violent neighborhood. Although school failureaffects more men, when it comes to the effect of victimization and fear of crime on school failure, womenare more affected than men.

6 ConclusionsThe objective of this study was to analyze the effects of fear of crime and violence on the progression of

adolescents and youth in the school system for Brazil, considering the PNAD information for the year 2009.The results of the study were generally statistically significant which indicates that analyze the negativeeffects of violence and fear of crime in school performance is relevant, once these variables can impair thelikelihood of adequate progress in educational system. It was found that the probability of failure in schoolis influenced by violence, mostly by physical assault. This result corroborates the literature that says that,because it involves physical contact and certain severity, has major consequences on the individual. Othertypes of crimes covered by this article also showed positive results to affect the likelihood of failure inschool, however, in smaller proportions of the physical assault.

Estimates for fear of crime allowed to infer that the fear in the neighborhood positively affects theprobability of school failure. This result is expected because fear in this level can represent fear in theschool environment since individuals generally studying in schools located in their neighborhoods, thusaffecting the performance of adolescents and youth in the school environment.

A characteristic that should be raised is that there is social inequality in the effects of victimization andfear of crime. The assumption has been that the effects of victimization and fear of crime on school failureprobabilities are more perceived by individuals with lower wealth. Families with lower purchasing powerhas limited mechanisms to prevent violence and fear and are, therefore, hardest hit.

It should be noticed that sex differences are most evident in the case of the variables related to personalcharacteristics. Work in school period has a negative effect among men, which was not observed amongwomen. For women, the variable pregnancy in school period proves important in affecting the likelihoodof school failure. In the case of fear of crime and physical assault, women are more sensitive than men are.

The results here obtained suggest that, in general, in addition to fear of crime and victimization, livingin the North, men, non-white, worked or became pregnant in school period, having a mother with loweducation, living in single-parent families and living in a larger family are characteristics that increase theprobability of being in school failure in Brazil.

Finally, the results suggest that public policies can play an important role when implemented consideringthe violence and fear in school environments, as well as in the neighborhoods, since these variables canpositively affect the likelihood of failure in school of Brazilian students. Another point that should be notedby policymakers is the most vulnerable to this violence and fear. Failure probabilities of the less wealthyare greater when observing these variables, showing a difference of effects, thereby this population demandmore attention in public policy.11 Lipschitz et al. (2000) stated that one of posttraumatic stress symptom that girls (aged, 12-21 years) can present is poorer

school performance.

15

Referências

ABRAMOVAY, M.; AVANCINI, M.; OLIVEIRA, H. O bê-á-bá da intolerância e da discriminação.Brası́lia, DF: Unicef, 2002.

ALBERNAZ, Â.; FERREIRA, F. H.; FRANCO, C. Qualidade e eqüidade na educação fundamentalbrasileiro. Pesquisa e Planejamento Econômico, v. 33, n. 3, 2002.

AQUINO, J. M.; PAZELLO, E. T. Trabalho materno e desempenho educacional das crianças: uma análiseda probabilidade de aprovação escolar. Instituto de Pesquisa Econômica Aplicada (Ipea), 2011.

BOWEN, N. K.; BOWEN, G. L. Effects of crime and violence in neighborhoods and schools on the schoolbehavior and performance of adolescents. Journal of Adolescent Research, Sage Publications, v. 14, n. 3,p. 319–342, 1999.

CAIRNS, R. B.; CAIRNS, B. D.; NECKERMAN, H. J. Early school dropout: Configurations anddeterminants. Child development, JSTOR, p. 1437–1452, 1989.

CAMERON, A. C.; TRIVEDI, P. K. Microeconometrics: methods and applications. [S.l.]: Cambridgeuniversity press, 2005.

CURRIE, J.; MORETTI, E. Mother’s education and the intergenerational transmission of human capital:Evidence from college openings. The Quarterly Journal of Economics, JSTOR, p. 1495–1532, 2003.

DELANEY-BLACK, V. et al. Violence exposure, trauma, and iq and/or reading deficits among urbanchildren. Archives of pediatrics & adolescent medicine, American Medical Association, v. 156, n. 3, p.280–285, 2002.

FERRARO, K. F. Fear of crime: Interpreting victimization risk. [S.l.]: SUNY press, 1995.

GAMA, V. A.; SCORZAFAVE, L. G. Os efeitos da criminalidade sobre a proficiência escolar no ensinofundamental no municı́pio de são paulo. Instituto de Pesquisa Econômica Aplicada (Ipea), 2013.

GOMES-NETO, J. B.; HANUSHEK, E. A. Causes and consequences of grade repetition: Evidence frombrazil. Economic Development and Cultural Change, JSTOR, v. 43, n. 1, p. 117–148, 1994.

GRISSOM, J. B.; SHEPARD, L. A. Structural equation modeling of retention and overage effects ondropping out of school. ERIC, 1989.

GROGGER, J. Local violence and educational attainment. Journal of human resources, JSTOR, p.659–682, 1997.

GUIMARÃES, E. Escola, galera e narcotráfico. Editora da UFRJ, 1998.

HALE, C. Fear of crime: A review of the literature. International review of Victimology, Sage Publications,v. 4, n. 2, p. 79–150, 1996.

HAVEMAN, R.; WOLFE, B. The determinants of children’s attainments: A review of methods andfindings. Journal of economic literature, JSTOR, v. 33, n. 4, p. 1829–1878, 1995.

HENRICH, C. C. et al. The association of community violence exposure with middle-school achievement:A prospective study. Journal of applied developmental psychology, Elsevier, v. 25, n. 3, p. 327–348, 2004.

HONDA, K. M. Um estudo sobre os determinantes do fracasso escolar. Dissertação (Mestrado) —Ribeirão Preto (SP): Universidade de São Paulo. Programa de Pós-Graduação em Economia, 2007.

16

KREIN, S. F.; BELLER, A. H. Educational attainment of children from single-parent families: Differencesby exposure, gender, and race. Demography, Springer, v. 25, n. 2, p. 221–234, 1988.

LAB, S. P.; WHITEHEAD, J. T. The school environment and school crime: Causes and consequences.1992.

LEON, F. L. L. d.; MENEZES-FILHO, N. A. Reprovação, avanço e evasão escolar no brasil. Instituto dePesquisa Econômica Aplicada (Ipea), 2002.

LIPSCHITZ, D. S. et al. Clinical and functional correlates of posttraumatic stress disorder in urbanadolescent girls at a primary care clinic. Journal of the American Academy of Child & AdolescentPsychiatry, Elsevier, v. 39, n. 9, p. 1104–1111, 2000.

MACHADO, D. C. Uma análise da freqüência e do fracasso escolar das crianças brasileiras. Sinais Sociais,Serviço Social do Comércio Administração Nacional, v. 1, n. 3, p. 36–65, 2007.

MACHADO, D. C.; GONZAGA, G. O impacto dos fatores familiares sobre a defasagem idade-série decrianças no brasil. Revista Brasileira de Economia, SciELO Brasil, v. 61, n. 4, p. 449–476, 2007.

MALTA, D. C. et al. Vivência de violência entre escolares brasileiros: resultados da pesquisa nacional desaúde do escolar (pense). Ciênc Saúde Coletiva, SciELO Brasil, v. 15, n. 2, p. 3053–63, 2010.

MILAM, A.; FURR-HOLDEN, C.; LEAF, P. Perceived school and neighborhood safety, neighborhoodviolence and academic achievement in urban school children. The Urban Review, Springer, v. 42, n. 5, p.458–467, 2010.

MOORE, K. A.; WAITE, L. J. Early childbearing and educational attainment. Family planningperspectives, JSTOR, p. 220–225, 1977.

MUDEGE, N. N.; ZULU, E. M.; IZUGBARA, C. How insecurity impacts on school attendance and schooldrop out among urban slum children in nairobi. International Journal of Conflict and Violence, v. 2, n. 1,p. 98–112, 2008.

PEARSON, F. S.; TOBY, J. Fear of school-related predatory crime. Sociology and Social Research,Proquest Social Sciences Journals, v. 75, n. 3, p. 117–125, 1991.

PENSE, I. B. d. G. e. E. Pesquisa Nacional de Saúde Escolar (PeNSE). 2012. Disponı́vel em:〈http://biblioteca.ibge.gov.br/visualizacao/livros/liv64436.pdf〉.

PONTILI, R. M.; KASSOUF, A. L. Fatores que afetam a freqüência e o atraso escolar, nos meios urbano erural, de são paulo e pernambuco. Revista de Economia e Sociologia Rural, SciELO Brasil, v. 45, n. 1, p.27–47, 2007.

PSACHAROPOULOS, G. Child labor versus educational attainment some evidence from latin america.Journal of population economics, Springer, v. 10, n. 4, p. 377–386, 1997.

SEVERNINI, E. R. A relação entre violência nas escolas e proficiência dos alunos. Catholic University ofRio de Janeiro, Rio de Janeiro, Brazil. Unpublished manuscript, 2007.

SKOGAN, W. G.; MAXFIELD, M. G. Coping with crime: Individual and neighborhood reactions. SagePublications Beverly Hills, CA, 1981.

TEIXEIRA, E. C.; KASSOUF, L. A. A relação entre violência nas escolas e desempenho escolar no estadode são paulo em 2007: uma análise multinı́vel. Rede de Economia Aplicada, WorkingPaper, n. 9, 2011.

TRIACA, L. M.; TEJADA, C. A. O. Determinantes do fracasso escolar: uma análise do ensino médio parao brasil. In: SUL, . XVII Encontro de Economia da R. (Ed.). XVII Encontro de Economia da Região Sul -ANPEC SUL 2014. [S.l.: s.n.], 2014.

17

http://biblioteca.ibge.gov.br/visualizacao/livros/liv64436.pdf

APPENDIX A

Table A1 – Summary statistics: school failure, characteristics, fear ofcrime and victimization, Brazil, 2009

Variables Average (%)Average with

school failure (%)Average without

school failure (%)Dependent variableDropout 16.62 41.83 -Repetition 22.44 56.48 -Never attended 0.68 1.70 -School failure 39.74 100.00 -Geographic characteristicsNorth 9.38 12.04 7.63Northeast 30.91 39.84 25.02South 13.85 10.86 15.81Southeast 38.54 30.62 43.76Midwest 7.32 6.64 7.77Personal characteristicsNon-White 55.33 66.35 48.07Man 50.84 57.76 46.27Worked in School Period 23.04 30.54 18.09Pregnancy in School Period 2.38 4.71 0.84Family characteristicsIncome 482.29 293.16 607.02Mother’s education 7.91 5.76 9.08Fatherless 22.92 26.88 20.75Number of Siblings 1.97 2.03 1.94Fear of Crime and VictimizationDomicile 19.22 19.57 18.99Neighborhood 30.60 29.01 31.65Municipality 43.45 39.89 45.80Attempted Robbery/Theft 1.29 1.00 1.48Theft 1.74 1.43 1.94Robbery 1.47 1.38 1.53Physical assault 1.46 2.02 1.09Multiple Crimes 3.23 2.79 3.51Estimated population 45,848,470 18,219,954 27,628,516

Source: PNAD data. Own Elaboration.

18

Table A2 – Estimation of variables and residual of mother’s education and income

Variables Mother’s education IncomeCoefficients SE Coefficients SEdependent variables Mother’s education Income

Constant 6.4384** 0.1883 315.9442** 33.2554Geographic characteristicsNortheast -0.9161** 0.0522 -46.7915** 9.2220South -1.6678** 0.0621 -104.1343** 10.9732Southeast -1.4070** 0.0533 -9.2330** 9.4040Midwest -0.7560** 0.0697 30.7802* 12.3156Personal characteristicsNon-White -0.6040** 0.0300 -96.9561** 5.2983Man 0.0432 0.0275 9.6037* 4.8537Worked in School Period -0.9961** 0.0343 13.2454* 6.0555Pregnancy in School Period -1.0142** 0.1313 -85.8482** 23.1799Family characteristicsFatherless 0.4236** 0.0332 10.2319* 5.8547Number of Siblings -0.3278** 0.0099 -40.9252** 1.7499Fear of CrimeDomicile -0.2602** 0.0462 -55.0618** 8.1559Neighborhood -0.0003 0.0461 -13.2145 8.1318Municipality 0.2816** 0.0366 45.8991** 6.4617VictimizationAttempted Robbery/Theft 0.4050** 0.1265 71.4876* 22.3394Theft 0.4262** 0.1044 49.6410* 18.4311Robbery 0.3421** 0.1163 77.2277** 20.5317Physical Assault -0.0853 0.1133 -15.7607 20.0018Multiple Crimes 0.3962** 0.0795 77.8917** 14.0379Instruments***Possession of Movable Property 0.9528** 0.0091 125.5344** 1.6008Residence Condition -0.1758** 0.0327 25.5887** 5.7792Locomotion 1.1000** 0.0351 264.2164** 6.1991Sewage / Trash / Water 1.1133** 0.0597 -86.3670** 10.5420Raw Material of the House -0.9860** 0.1811 -111.9703** 31.9780

R2 0.3480 Prob>F = 0.000 R2 = 0.2533 Prob>F = 0.000

Note: control variables omitted from the table; p-value

Table A3 – Endogeneity test for the variables mother’s education and income

Variables Mother’s education Income***Coeficientes SE Coeficientes SEdependent variables School failure School failure

Constant 0.8007** 0.0115 0.0000** 0.0000Geographic characteristicsNortheast -0.0606** 0.0062 -0.0335** 0.0062South -0.1305** 0.0072 -0.0976** 0.0072Southeast -0.1099** 0.0061 -0.0870** 0.0061Midwest -0.0881** 0.0081 -0.0870** 0.0082Personal characteristicsNon-White 0.0023 0.0037 0.0006 0.0037Man 0.1338** 0.0032 0.1346** 0.0032Worked in School Period 0.0215** 0.0042 0.0723** 0.0040Pregnancy in School Period 0.3069** 0.0155 0.3244** 0.0155Family characteristicsIncome -0.0000 0.0000 0.0003** 0.0000Mother’s education -0.0619** 0.0009 -0.0212** 0.0004Fatherless 0.0732** 0.0038 0.0509** 0.0038Number of Siblings 0.0123** 0.0012 0.0164** 0.0012Fear of CrimeDomicile -0.0189* 0.0055 -0.0221** 0.0055Neighborhood 0.0158* 0.0054 0.0104 0.0054Neighborhood -0.0067 0.0043 -0.0091* 0.0043VictimizationAttempted Robbery/Theft 0.0047 0.0149 0.0027 0.0149Theft 0.0156 0.0123 0.0053 0.0123Robbery 0.0081 0.0137 0.0127 0.0137Physical assault 0.0912** 0.0133 0.0881** 0.0133Multiple Crimes 0.0234* 0.0094 0.0231* 0.0094ResidualsMother’s education -0.0203** 0.0004 Income 0.0000 0.0000

R2=0.2122 Prob>F = 0.0000 R2=0.2101 Prob>F = 0.0000

Note: control variables omitted from the table; p-value