Armed attacks, child labor and school attendance

Ousmane Mariko

Faculty of Economics and Management, Bamako, Mali, Associate Researcher at the

Grenoble Economic Research Center (CREG), France.

Mohamed Niaré

Faculty of Economics and Management, Bamako. Email: mniare1411@[Link]

Moustapha Dembélé

Beijing Technology and Business University (BTBU), Beijing, China.

Abstract: In this paper, we estimate the effect of armed attacks on the economic work and school
attendance of children aged 5-14 in Mali. The study uses microeconomic data from Modular and
Permanent Household Survey MPHS (2019) and Armed Conflict Location and Event Data
ACLED (2019). The results show a positive impact of armed clashes on children's economic labor,
while the effect is negative on school attendance. Econometric analysis also confirms that factors
that contribute to increasing child labor simultaneously reduce school attendance. In addition, we
have identified some possible transmission channels to explain the results found. This analysis
shows that the number of armed attacks exacerbates child labor, on the one hand by lowering the
average income of the administrative subdivision, and on the other by reducing household income.
This analysis thus confirms that child labor is not only linked to the household, but also to the
overall economic situation. Finally, the results obtained from Fairlie's (2005) decomposition show
that the security situation helps to exacerbate the rural nature of child labor, explaining around
53% of the total gap explained. The results of this paper highlight strong evidence of the impact
that conflict can have on the exacerbation of child labor and the accumulation of human capital.
Consequently, public authorities will need to take urgent action to protect children in conflict
zones, particularly those in rural areas.

Keywords: armed conflict, household survey, school attendance, child labor.

JEL code: C25, D91, J01

1
Introduction

The protection of children in areas of armed conflict is a crucial issue in affected countries and has
been the focus of particular attention from political decision-makers for several decades.
International humanitarian law grants special protection to children during conflicts, through the
four Geneva Conventions of 1949 and their Additional Protocols of 1977. These provisions cover
access to food and medical care, education, detention, separation from family and participation in
combat. The International Convention on the Rights of the Child (ICRC) also calls on states to
facilitate the psychological and physical recovery and social integration of children affected by
conflict. In addition, ILO Convention No. 182 on the Worst Forms of Child Labor, which was
adopted in 1999 and came into force in 2000, commits all States to eradicating the recruitment of
children in conflicts.

Since the establishment of these international conventions for the protection of children, Mali has
ratified most of these legal instruments. In addition, violence against children is covered by the
constitution, laws and regulations, and customary law. Despite these efforts, the country faces a
huge challenge in protecting children in conflict zones. Indeed, the Global Protection Cluster report
(2018) reported incidents of slavery in regions plagued by insecurity, where children were sold to
perform agricultural and domestic work by their communities. Armed groups also put children to
work in gold mines and use the profits to enrich themselves, fuel the arms trade and finance
violence (United Nations High Commissioner for Refugees, 2020). In addition, as a result of the

2
conflict, 215 children have been forcibly recruited by armed groups to fight, including as suicide
bombers (Global Protection Cluster, 2020).

Beyond this physical violence against children, armed groups also attack school infrastructures,
destroying them or using them as headquarters. Indeed, due to jihadist and community clashes,
more than 1,113 schools were closed in 2019 in Mali, affecting around 334,000 children (UNICEF,
2019). This situation can lead parents to put their children to work to the detriment of their
schooling. Finally, armed conflict can also lead to indirect economic consequences through the
destruction of production infrastructure, displacement of populations, disruption of economic
activities, etc., which are likely to influence household decisions regarding child labor and early
school dropout. Consequently, by preventing children from attending school, armed conflicts are
likely to trap households in poverty. Indeed, children who have dropped out of school are less
likely to apply for better-paid formal employment, as they have not accumulated a sufficient level
of human capital.

A number of studies have analyzed the impact of armed conflict on children. Most have focused
on how conflict affects school enrolment (Chamarbagwala and Morán, 2011; Leon, 2012;
Rodriguez and Sanchez, 2012, etc.) and educational outcomes (Justino, and al, 2013; Brück and
al, 2013; Bertoni and al, 2018, etc.). The economic literature on the impact of armed conflict on
child labor remains limited, however. Indeed, it is only recently that some empirical work
(Rodriguez and Sanchez, 2012; Di Maio and Nandi, 2013; Naufal and al, 2019) has highlighted
that armed conflict matters in explaining child labor. In this context, Mali is a particularly
interesting case, given its security situation and the weight of children in its population. Indeed,
according to ACLED (2019) data, in 2019 alone the country recorded more than 76 armed clashes
between politically organized armed groups. What's more, among the ten least peaceful African
countries (Institute for Economics and Peace, 2020), Mali has the highest proportion of children
under 15 in its population (47.3%) (World Bank Population Division, 2020).

Based on the above, the aim of this article is to analyze the impact of armed attacks on the work
and school attendance of children aged 5 to 14 in Mali. This study contributes to the empirical
literature by providing the first empirical evidence of the impact of armed conflict on child labor
in Mali, while the very limited number of existing studies on the subject focus on countries in

3
different geographical areas (Colombia, Israel, Iraq). As such, our findings are more likely to be
useful for other African countries facing difficult security situations.

The rest of the article is divided into three sections. The first section reviews the theoretical and
empirical literature. The second section presents the data and the empirical strategy. Finally, the
third section presents and discusses the empirical results.

1. Literature review
From a theoretical point of view, although there are no studies that have directly assessed the effect
of armed conflict on child labor, some work could nevertheless help identify potential transmission
channels between the two concepts. The first channel relates to negative exogenous shocks to
household income. Indeed, studies show that unexpected negative events prompt parents to make
children work to protect themselves against the risks of interrupted resource flows (Beegle, 2005;
Lachaud, 2008; Bandara and al, 2015). Since armed conflicts are assimilated to negative
exogenous shocks, they are likely to lead parents to resort to child labor as an additional source of
income in order to smooth family consumption. As for the second channel, it concerns the theory
of demand for education initiated by Becker (1975). This theory states that the decision to send
children to school is the result of a trade-off between the return associated with education and its
cost. Thus, when the return on schooling is low or uncertain, parents perceive the investment in
education as a loss of income and choose to send their children to work. Consequently, in conflict
zones where the quality of education is likely to be low due to teacher attrition, the use of school
infrastructure for military purposes, a drop in educational spending in favor of defense, etc.,
parents may feel that it is preferable for their children to drop out of school and work, since
education is no longer perceived as a guarantee of access to more productive and better-paid jobs.
The third transmission channel is linked to recent literature suggesting a positive relationship
between life expectancy and human capital accumulation (Soares, 2005; Estevan and Baland,

4
2007). In fact, these studies show that a high mortality rate among young people can lead parents
to consider the maturation of educational investment as uncertain, prompting them to resort to
child labor. In this context, armed conflicts can alter children's life expectancy (recruitment of child
soldiers, increase in physical violence, deterioration in sanitary conditions, etc.), reduce investment
in education, and increase participation in the labor market.

Empirically, the most important work focuses on the impact of armed conflict on children's
education. Conflict is associated with low levels of schooling (Chamarbagwala and Morán, 2011;
Leon, 2012; Rodriguez and Sánchez, 2012) and a reduction in the quality of education (Justino,
and al, 2013; Brück and al, 2013; Bertoni and al, 2018). However, few studies have analyzed the
impact of armed conflict on child labor. Using Colombian data, Rodriguez and Sanchez (2012)
find that armed conflict measured by the number of attacks on infrastructure, civilians and state
security forces, pushes children to drop out of school and take part in the labor market. In a similar
vein, Di Maio and Nandi (2013) analyze the impact of the Israeli-Palestinian conflict on the work
and school attendance of children in the West Bank. The authors use the number of days the border
between the two states is closed to measure the intensity of the conflict. They conclude that the
escalation of clashes increases the labor supply of Palestinian children and reduces their school
attendance. The study also identified possible transmission channels to explain the observed
results. Among these, the authors demonstrated that the number of days the border is closed
contributes to increased child labor by worsening labor market conditions in the West Bank and
reducing the likelihood of the father being employed in Israel. Finally, another study by Naufal
and al. (2019) in Iraq examines the relationship between armed clashes, assessed by the number
of fatalities, and child labor. They highlight that the intensification of clashes increases the risk of
a child engaging in economic activity, but does not affect his or her involvement in domestic tasks.

2. Data and estimation strategy

For the purposes of this article, the sample is limited to children aged 5-14. There are two reasons
for this choice. Firstly, the International Labor Organization (ILO) uses the age of 5 as a starting
point for measuring child labor, as this corresponds to the start of formal education in many
countries. It is therefore important to ensure that these children are not forced to work. Secondly,
the upper limit of 14 refers to ILO Convention 138, which sets 15 as the minimum age for
admission to employment.

5
The data used in this study were obtained from two sources. Firstly, information on the
characteristics of the child, the head of household and the household was collected from the
Modular and Permanent Household Survey (MPHS, 2019). This is a two-stage stratified sample
survey with proportional allocation in the first stage. It covered 6,634 households, totaling 28,943
individuals spread over four passages. We merge this dataset with that of ACLED (2019), which
records conflict events. Indeed, the ACLED database collects information on armed conflict,
protests and political violence from local, national and international sources, including the media,
verified social media accounts, reports from governments, NGOs and partner organizations.

In this study, we use probit models to analyze the effect of armed conflict on child labor and
schooling. Child labor corresponds to the child's status on the labor market. It is coded "1" if the
child is in paid employment, and "0" otherwise. We define paid activity as any activity performed
by the child for at least one hour in return for income during the reference week. School attendance
refers to children who were enrolled in school and attended school during the reference week. The
data in Table 1 show that around 11% of children aged 5 to 14 participate in the labor market,
while 55% attend school.

Table 1: Socio-demographic characteristics of the sample

Observations Mean Std Dev Min Max
Child labor* 8,304 0,11 - - -
School attendance * 8,304 0,55 - - -
Boy* 8,304 0,52 - - -
Child of head of household 8,304 0,69 - - -
Household size 8,304 12,75 8,25 2 52
Educated head of household* 8,304 0,27 - - -
Male head of household* 8,304 0,96 - - -
Households with access to bank credit* 8,304 0,05 - - -
Rural* 8,304 0,60 - - -
Source: MPHS 2019. * Indicates a categorical variable
The probability models are as follows:

P (𝐭𝐫𝐚𝐯 = 𝟏)= 𝛃𝟎 + 𝛃𝟏 𝐥𝐧𝐜𝐨𝐧𝐟𝐥𝐢𝐭 + 𝛃𝐤 𝐗 𝐤 + 𝐅𝐄 + 𝛍𝐢 (1)

P (𝐒𝐜𝐨𝐥 = 𝟏)= 𝛂𝟎 + 𝛂𝟏 𝐥𝐧𝐜𝐨𝐧𝐟𝐥𝐢𝐭 + 𝛂𝐤 𝐗 𝐤 + 𝐅𝐄 + 𝛆𝐢 (2)

Where the dependent variable 𝐭𝐫𝐚𝐯 is equal to one if the child is in paid employment, while 𝐒𝐜𝐨𝐥
is equal to one if the child attends school. 𝐥𝐧𝐜𝐨𝐧𝐟𝐥𝐢𝐭 is the independent variable of interest in both

6
specifications, i.e. the logarithm of the number of armed clashes involving two politically
organized groups. To construct this variable, we take into account all battles occurring in Mali
during 2019. The Geographic Information System (GIS) border file showing Mali's administrative
units and borders at the second geographical level is obtained from Integrated Public Use
Microdata Series (IPUMS). We geolocate the battles on the map and calculate the number of
clashes per administrative region. We hypothesize that clashes generate political and economic
instability, increase child labor and reduce school attendance (Di Maio and Nandi, 2019). 𝐗 𝐤 is a
vector composed of the set of control variables, including the child's gender, the child's position in
the household, the gender of the household head, the household head's education, the household's
access to bank credit, household size and area of residence. 𝛃𝟎 and 𝛂𝟎 are the constants, 𝐅𝐄
represents the region fixed effect, while 𝛍𝐢 and 𝛆𝐢 are the error terms.

Table 2: Distribution of armed attacks by region

Region Observations Mean Std, Dev Min max

Kayes 5,416 0,284 0,451 0 1
Koulikoro 3,268 0,403 0,702 0 2
Sikasso 7,613 0,914 1,250 0 4
Ségou 4,243 1,811 1,594 1 5
Mopti 1,309 19,319 7,522 6 29
Timbuktu 1,888 21,746 4,044 5 23
Gao 280 10 0,000 10 10
Bamako 4,926 0,987 0,161 0 2
Source : ACLED 2019

3. Presentation and discussion of results

[Link] of the possible endogeneity of the independent variable of interest

This section analyzes the influence of armed attacks on the economic work and school attendance
of children aged 5 to 14. According to the Global Protection Cluster report (2020), 215 children
were forcibly recruited by armed groups to fight or work in the gold mines. Although we have no
evidence of the systematic use of child soldiers, our variable of interest presents a risk of
endogeneity if there is any pressure for them to leave school to join armed groups or engage in
economic activities to finance the violence. Finally, to exclude any risk of reverse causality, we
regress the logarithm of the number of attacks on children's economic work and school attendance.

7
The results in Table 3 confirm that in all specifications, the coefficients are not statistically different
from zero. Consequently, we can rule out the possibility that armed attacks are a response to
children's economic occupation or school drop-out.

Table n°3: Effect of children's work and school attendance on the number of armed clashes.

Log(attacks) Log(attacks)
0,13
Economical work
(0,21)

0,11
School attendance
(0,10)

Region fixed effects Yes Yes

0,84** 0,67*
Constant
(0,38) (0,33)

R2 0,15 0,13
Prob > F 0,09 0,001
Observations 5 310 3 153
Note: linear regression results. ***, **, * are significant at 1%, 5% and 10% levels respectively.

[Link] results

Table 4 shows the results of the probit estimations. Region fixed effects are included in the different
models to control for unobserved regional characteristics that might influence child labor and
schooling. Data in brackets represent standard errors. In addition, the magnitude of the coefficients
is measured by the marginal effect (not reported in the table). The results in columns (1) and (3)
indicate that increasing armed attacks by one unit increases the probability of child labor by 4%

8
and reduces school attendance by 2%. This result corroborates the findings of Rodriguez and
Sanchez (2012) in Colombia and Di Maio and Nandi (2013) in Palestine. Although it remains
difficult to identify potential transmission channels at this stage, the results nevertheless confirm
the need to take the security situation into account when explaining child labor and school
attendance, as has been documented in other parts of the world.

Table 4: Probit estimates of the effect of armed attacks on children's work and school
attendance.

Economical Economical School School

work work attendance attendance
0,26*** 0,23*** -0,15* -0,13*
Log (number of attacks)
(0,04) (0,05) (0,08) (0,08)
0,001 0,11
Child of head of household
(0,06) (0,10)
-0,18*** 0,11
Girl
(0,05) (0,08)
-0,20*** 0,23***
Educated head of household
(0,07) (0,09)
-0,51*** 0,45*
Female head of household
(0,19) (0,28)
0,001 0,44*
Household with access to credit
(0,15) (0,29)
0,002 0,01
Household size
(0,003) (0,01)
0,23*** -0,11
Rural
(0,06) (0,10)
Region fixed effects Yes Yes Yes Yes
-1,51*** -1,56*** 1,54*** 1,28***
Constant
(0,09 (0,13) (0,11) (0,18)
Observations 5 299 5 233 3 127 3 094
Note: probit regression results. ***, **, * are respectively significant at the 1%, 5% and 10% level.

9
 [Link] checks

To check the robustness of our results, we include the following control variables in columns (2)
and (4) of Table 4: gender of the child, gender of the head of household, level of education of the
head of household, household access to bank credit, household size and area of residence. We note
that the addition of these variables does not alter the significance of the variable of interest.
However, the size of the marginal effect on work and school attendance is slightly reduced by one
percentage point. The results are therefore robust to the control variables. Furthermore, analysis of
the inclusion variables shows that educated or female heads of household demonstrate greater
altruism towards children by reducing their participation in economic activities in favor of greater
school attendance. Moreover, being a boy or living in a rural area exposes children more to
employment. Another interesting result concerns the beneficial effect of household access to bank
credit on children's school attendance.

So far, we have opted for the simplest econometric approach, i.e. analyzing work and school
attendance using a univariate specification. However, this strategy has the disadvantage of not
taking into account the interdependence between children's work and schooling decisions. It is
therefore preferable to analyze the robustness of our results in a multivariate framework. Three
models are generally run in the literature and are based on divergent assumptions. Firstly, we have
the sequential probit, whose use assumes that the decision to allocate a child's time between work
and school is taken in an a priori determined order. However, the hierarchy of orders on which the
results of this model depend remains controversial (Mbebi, 2018). Indeed, some parents may feel
that their child's happiness depends on being gainfully employed, while for others, schooling
remains the best way to guarantee a better future for the child. The second model, the multinomial
logit, is based on the assumption of independence between decisions to work and to send children
to school. In most cases, this assumption is violated, as the household's choice is made by
simultaneously considering the advantages and disadvantages of each option. For these reasons,
the method chosen to assess the robustness of our basic results is bivariate probit, which allows us
to simultaneously control for the interdependence between working and schooling children.

The results of estimating the bivariate probit model show that, whatever the choice of specification,
the correlation coefficient " 𝜌 " between child labor and school attendance is negatively different
from zero at the statistical threshold of 1%. Thus, factors that contribute to increasing child labor

10
also reduce school attendance. As expected, armed attacks are positively related to child labor and
negatively related to school attendance. This once again confirms the robustness of the results
obtained. Furthermore, the other determinants of work and school attendance remain unchanged
from the baseline results, with the exception of the variable measuring whether the child lives with
his/her biological parents, which becomes significant in the work equation.

Table 5: Biprobit estimate of the effect of armed attacks on children's work and school
attendance.

Economical School Economical School

work attendance work attendance
0,20*** -0,12* 0,17** -0,11*
Log (number of attacks)
(0,07) (0,08) (0,07) (0,09)
-0,13* 0,14
Child of head of household
(0,09) (0,10)
-0,11* 0,09
Girl
(0,08) (0,09)
-0,10 0,22**
Educated head of household
(0,09) (0,09)
-0,72*** 0,44*
Female head of household
(0,27) (0,27)
-0,27 0,46*
Household with access to credit
(0,30) (0,30)
0,01 0,01*
Household size
(0,01) (0,01)
0,22** -0,13
Rural
(0,09) (0,10)
Region fixed effects Yes Yes
-1,97*** 1,56*** -1,98*** 1,28***
Constant
(0,14) (0,11) (0,19) (0,17)
Observations 3 187 3 153
𝜌 (Wald; sig) -0,78***(117,031 ; 0,000) -0,75*** ( 101,59 ; 0,000)
Note: biprobit regression results. ***, **, * are respectively significant at 1%, 5% and 10% level.

11
 [Link] channels

To identify the transmission channels of the effect of armed attacks on child labor, we first regress
the number of armed clashes on the average wage of the administrative subdivision and household
income. Column (1) of the table below shows that the increase in armed attacks significantly
reduces the average wage of the administrative subdivision where the child lives. Column (3) in
turn indicates that the drop in this average income leads to an increase in child labor. Thus, the
deterioration in the overall economic situation caused by the armed attacks contributes to an
increase in the economic occupation of children. Secondly, the results reported in columns (2) and
(4) confirm that the number of armed attacks also reduces household income, which in turn
exacerbates child labor. These results are in line with studies suggesting that child labor is a
household response to negative exogenous shocks (Guarcello andal, 2010; Bandara andal, 2015;
etc.). Indeed, unexpected negative events prompt parents to make children work to protect
themselves against the risks of interrupted resource flows.

Table 6: Transmission channels for the effect of armed attacks on work and school
attendance
Log (average income Economical Economical
Log (household
for administrative c d
income) b work work
subdivision) a
-0,13*** -0,17***
Log (attacks)
(0,002) (0,016)
Log (average income for -1,40***
administrative subdivision) (0,18)
-1,11***
Log (household income)
(0,25)
Region fixed effects Yes Yes Yes Yes
9,63*** 9,34*** 12,91*** 11,51***
Constant
(0,005) (0,04) (1,80) (2,47)
R2 0,52 0,10
Pseudo R2 0,10 0,48
Probability 0,000 0,000 0,000 0,000
Observations 19 131 4 210 8 304 335
a b c
Note: result of linear regression. result of linear regression. result of probit regression.
d
probit regression result. ***, **, * are respectively significant at the 1%, 5% and 10% level.

[Link] analysis

12
 According to MPHS (2019) data, around 81% of working children come from rural areas. This
descriptive finding is confirmed by our empirical results, which show that living in a rural
environment increases the probability that a child will engage in economic activity. With reference
to other studies, this result is not specific to Mali (Couralet, 2003; Haddad, 2017, etc.). To identify
the explanatory factors of this gap, we use Fairlie's (2005) non-linear decomposition applied to a
probit specification. This technique divides the total gap into two parts: explained and unexplained.
The explained part represents differences in the distribution of observable variables. The
unexplained part covers unobservable variables, notably cultural factors. It also measures the
discriminatory component of the total gap.

Empirical analysis shows that the model used explains 71% of the urban-rural disparity in
economic child labor. However, 29% of the total gap remains unexplained and can be attributed to
cultural norms that encourage child labor. Indeed, certain popular African beliefs consider child
labor to be a means of transmitting necessary skills, enabling the child to take over from the parents
as an adult. Furthermore, when analyzing the contribution of each independent variable, we find
that the security situation explains over 53% of the explained gap (38% of the total difference)
between urban and rural children in terms of labor market participation. It is also interesting to
note that household head characteristics, accounting for 32.29% of the explained difference, are
the second most important factor influencing the rural nature of child labor. This is followed by
household characteristics (9.33% of the variance explained) and child characteristics (5.22% of
the variance explained). These results show that the prevalence of child labor in rural areas can be
reduced by improving the security situation, making adults literate so that they realize the
importance of education, strengthening women's decision-making power within households, and
promoting nuclear families.

Table 7: Fairlie's (2005) breakdown of the child labor gap by place of residence
Coefficient Percentage
Total difference 0,10476967
Difference explained 0,07481615 71,41
Unexplained difference 0,02995352 28,59
Rural children: engaged in Economical work 15,6
Urban children: engaged in Economical work 5,12
Total deviation 10,48

13
 %
% total
Coefficient Standard error Z difference
difference
Contribution to difference explained
Log (attacks) 0,0399196*** 0,0084671 4,71 53,36 38,10
Child's position 0,0005784 0,0010921 0,53 0,77 0,55
Child's gender 0,0033283*** 0,0008876 3,75 4,45 3,18
Education of the head of
0,0207717*** 0,0027863 7,45 27,76 19,83
household
Gender of head of household 0,0033905*** 0,0007853 4,32 4,53 3,24
Household with access to credit 0,0005432 0,0004186 1,3 0,73 0,52
Household size 0,0064335*** 0,0018127 3,55 8,60 6,14
Source: authors' calculations based on MPHS (2019) and ACLED (2019)

Conclusion
This article uses data from MPHS (2019) and ACLED (2019) to estimate the influence of armed
attacks on child labor and school attendance among children aged 5-14 in Mali. We showed that
an increase in armed attacks increases child labor and reduces school attendance. This implies that
armed clashes have a detrimental impact on human capital accumulation. Furthermore, using a
bivariate probit to check the robustness of our results, we found that work and school attendance
are interdependent, but are differentially affected by the security situation in terms of significance
and magnitude. The study also identified some possible transmission channels to explain the
results. We demonstrated that the number of armed attacks reduces the average wage of the
administrative subdivision and household income. These two factors are also associated with an
increase in child labor. This analysis confirms that child labor is not only linked to the household,
but also to the overall economic situation. Finally, the results obtained from Fairlie's (2005)
decomposition show that the security situation also contributes to exacerbating the rural nature of
child labor, explaining around 53% of the total gap explained. The results of this article thus
provide strong evidence of the impact that conflict can have on the exacerbation of child labor and
the accumulation of human capital. Consequently, public authorities will need to take urgent action
to protect children in conflict zones, particularly those in rural areas.

14
References
Bandara, A., Dehejia, R.H, & Lavie-Rouse, S. (2015). "The Impact of Income and Non-income
Shocks on Child Labor: Evidence from a Panel Survey of Tanzania. " World Development,
67, pp. 218-237.

Becker, G. S. (1975). "Human capital and the personal distribution of income: An analytical
approach". In G. S. Becker (Ed.), Human capital: A theoretical and empirical analysis (2nd
ed., pp. 94-144). Columbia University Press.

Beegle, K., Dehejia, R.H., & Gatti, R. (2006). "Child Labor and Agricultural Shocks". Journal of
Development Economics, 81(1), pp. 80-96.

Bertoni, E., Di Maio, M., Molini, V., & Nistico, R. (2019). "Education is forbidden: The effect of
the Boko Haram conflict on education in North-East Nigeria". Journal of Development
Economics, 141, p. 102249.

Brück, T., Di Maio, M., & Miaari, S. (2013). Exposure to violence and student achievement in
palestine: Evidence from the second Intifada. DiW Berlin Working Paper, Berlin, Germany.

Chamarbagwala, R., & Morán, H.E. (2011). "The human capital consequences of civil war:
Evidence from Guatemala". Journal of Development Economics, 94(1), pp. 41-61.

15
Couralet, P-E. (2003). Child labor in Latin America. OECD Working Papers No. 3: Social Issues,
Employment and Migration.

Di Maio, M., & Nandi, T. (2013). "The effect of the Israeli-Palestinian conflict on child labor and
school attendance in the west bank". Journal of Development Economics, 100, p.107-116

Estevan, F. & Baland, J.-M. (2007). "Mortality risks, education, and child labor". Journal of
Development Economics, 84(1), 118-137.

Fairlie, R. W. (2005). "An extension of the Blinder-Oaxaca decomposition technique to logit and
probit models". Journal of Economic and Social Measurement, 30(4), pp. 305-316.

Global Protection Cluster (2018). Child Protection Risks and Needs in Mali, Children Resource,
UNHCR, Geneva.

Global Protection Cluster (2020). Aftershock: abuse, exploitation, and human trafficking in the
wake of covid-19, UNHCR, Geneva.

Guarcello, L., Mealli, F., & Rosati, F. C. (2010). "Household Vulnerability and Child Labor: The
Effects of Shocks, Credit Rationing, and Insurance". Journal of Population Economics,
23(1), pp. 69-198.

Haddad, G.K. (2017). "Parents' decision on child labour and school attendance: evidence from
Iranian households". Journal of Education and Work, pp. 1-21.

Institute for Economics and Peace (2020). Global Peace Index 2020, IEP, Sydney.

Justino, P., Marinella, L., & Paola, S. (2013). "Short- and long-term impact of violence on
education: The case of Timor Leste". World Bank Economic Review, 28(2), pp. 320-353.

Lachaud, J.P. (2008). "Child labor and poverty in Africa: a re-examination applied to Burkina
Faso". Economics and Forecasting, 5(186), p. 47-65.

Leon, G. (2012). "Civil conflict and human capital accumulation: The long-term effects of political
violence in Peru". The Journal of Human Resources, 47(4), pp. 991- 1022.

Mbebi, O.E. (2018). "The determinants of child labor in Cameroon: the role of residential
environment and gender". Development Economics Review, 26, pp. 5-52.

16
Naufal, G, Malcolm, M., and Diwakar, V. (2019). "Armed conflict and child labor: evidence from
Iraq". Middle East Development Journal, pp. 1-18.

Rodriguez, C., and Sanchez, F. (2012). "Armed conflict exposure, human capital investments, and
child labor: Evidence from Colombia ". Defense and Peace Economics, 23(2), pp. 161-184.

Soares, R.R. (2005). "Mortality Reductions, Educational attainment, and Fertility Choice ".
American Economic review, 95(3), pp. 580-601.

UNICEF (2019). Sharp increase in grave violations against children in Mali, Warns UNICEF,
UNICEF Mali.

World Bank (2020). World Bank estimates based on age/sex distributions of United Nations,
Population Division's, Washington DC.

17