Decent Job or Migration: What is the Best Strategy to Fight
against Poverty and to Promote Children’s Education in
Senegal?
Ababacar S. GUEYE *
CERDI, University of Auvergne, France1
Preliminary Version
Abstract
In this paper we try to answer a simple normative question: between migration and
access to a local decent job, which is the more effective mean to reduce poverty? Some
studies have analyzed the relation between labor market and poverty, and a large
literature has demonstrated the positive impact of migration on poverty. Yet, to the best
of our knowledge, no studies have attempted to compare explicitly the effects of a good
local job and migration on poverty. Does a decent job represent a good substitute for
migration? To answer this question, we use data from the 2011 Senegalese Survey of
Monitoring Poverty (ESPS 2011) and an instrumental variable strategy to correct for
endogeneity bias due to self-selection in migration and access to labor market. Proxies of
migration and labor market networks captured from ethnicity and geographic location
are used as instruments. Our results confirm the economic literature on the positive
impact of migration on poverty, but show that access to decent work has a much greater
impact on poverty even if we just consider migration to developed countries. We then try
to understand what channel underlies this result. Does a decent job allow to look to the
future and to invest more in children’s human capital? We find that decent job promotes
children’s schooling whereas we find no significant effect of migration in schooling. The
results are robust to different specifications of the model and different measures of
poverty and decent job.
Keywords: poverty, migration, decent work, education, instrumental variable, network
* E-mail : [email protected]
1 CERDI : Centre d’Etudes et de Recherches sur le Développement International, 65 Boulevard François Mitterrand,
63000 Clermont-Ferrand, France
1
Introduction
Senegal is one of the poorest countries in the world ranked 170 out of 188 countries by
the 2014 Human Development Index. Labor market issue is known as one of the main
problems of the country. The unemployment rate according to the ILO definition is
estimated about 10.3% (ANSD, 2013), well above the overall average in Sub-Saharan
Africa which stands for 7.6%. The economic system produces so few jobs for the rising
young population entering the labor market. Furthermore the unemployment rate is only
partially informative of the functioning of the labor market. The precariousness of
employment is a major concern, 32% of workers are underemployed (ANSD, 2013) and
87% are employed in the informal sector. The 2013 census data indicates that 60% of
unemployed are aged between 15 and 34 (ANSD, 2016). This situation makes migration
a valuable alternative for young people. Thereby, many young people opt to migrate to try
to lead a better life beyond the borders of their country. Even when migration policies are
more restrictive, young are moving towards illegal roads to reach Europe. Young
Senegalese were particularly involved in the waves of illegal migration in the early and
mid-2000s. Mbaye (2014) indicates that, in 2006, half of the 30,000 illegal migrants
landed in the Canary Islands were Senegalese and argues that illegal migrants overstate
the returns to migration. These aspects lead us to wonder about the real values of the
returns to migration. In this paper I attempt to analyze which alternative between
migration and access to a decent local job has greater impact on poverty reduction and
human capital investments.
The positive impact of migration on poverty is a well-known fact as it was so widely
demonstrated in the economic literature in different parts of the world: Acosta and al.
(2008) in Latin America, Gupta and al. (2009) in Sub-Saharan Africa, Imai and al. (2012)
in Asia and the Pacific etc. In West Africa in particular, micro studies find substantial
positive impacts of migration. Lachaud (1999) shows a 14.9% reduction effect of
remittances on poverty in rural Burkina Faso. Gubert et al. (2010) in the case of Mali find
that remittances reduce poverty by 11%. In Nigeria, Chiwuzulum Odozi et al. (2010) find
that remittances decrease poverty by 20%.
With its large diaspora, Senegal holds a wide stock of remittances which constitute an
important source of income. Indeed, the share of remittances in GDP is about 11% (World
2
Bank, 2015) placing Senegal at the fifth position in Africa after Gambia, Leshoto, Liberia
and Comoros.
On the other side, there is a strong link between labor market status and poverty. Having
a good job is definitely one of the main drivers of poverty eradication. According to
Salazar-Xirinachs Executive Director in the International Labor Office “having access to
stable and protected employment is the most sustainable path to exiting poverty and
promoting inclusion” (Cazes and Verick, 2013). Gutierrez et al. (2007) find that creating
intensive jobs in the secondary sector and productive employment in agriculture lead to
poverty reduction. Ernst and Berg (2009) argues that a raise in productive and
remunerative employment increases poor’s incomes which leads to a reduction of
poverty. Beyond the pecuniary aspect, a decent job contributes to reduce poverty through
the stability of labor income, the security and the protection of employment. In fact a
secure job, even if it is not very highly paid, allows people to manage smartly their income
and to invest more in their family well-being. Banerjee and Duflo (2011) argues that when
a member of the family gets a stable job, schools are more likely to accept his children,
hospitals are more likely to provide high quality and more expensive cares for sick family
members and the other relatives may be able to invest more adequately to develop their
business.
To the best of my knowledge this is the first study which attempts to compare explicitly
the impact of access to decent job and migration on poverty. Some papers rely on the
seminal work of Barham and Boucher (1998) to create a counterfactual scenario in which
migrants would be employed in local jobs. Margolis and al. (2015) in a study conducted in
two towns in Algeria show that even in the more optimistic scenario where all migrants
would have a formal work, migration would still has a significant impact on poverty.
However these studies cannot account all the general equilibrium effects that will result
to a massive integration of migrants in the local labor market. My empirical strategy is
different and analyzes more specifically the impact of access to decent employment.
Basically, the theoretical idea is to consider migration and access to decent employment
as two alternatives available to households, and the purpose is just to analyze which one
3
benefits more to the household in terms of poverty reduction and improvement of living
standards.
To answer this question we use the Senegalese Survey of Monitoring Poverty (ESPS 2011)
conducted in 2011 by the National Agency of Statistics and Demography (ANSD) and
construct an index of decent employment using variables related to social protection and
stability of employment. Obviously endogeneity problems are very likely to arise and bias
the estimates. In fact people are self-selected in the access to decent employment and
migration as well. Not considering this issue can lead to an overestimation of the effect of
migration and decent work. Controlling this endogeneity problem is often a tricky
econometric challenge. We rely on an instrumental variable strategy to overcome selfselection due to migration and access to decent work. The instruments proxy the network
of decent employment and migration using geographical proximity and membership to
ethnic group. I find significant and positive impacts on poverty reduction both for
migration and access to decent employment but the latter have higher effects on poverty
even when restricting migration to developed countries.
Following the reasoning by Banerjee and Duflo (2011), I test also the impact of these
alternatives on human capital investments and particularly children schooling. Results
show a high impact of decent employment on children’s enrolment but migration has no
significant effect once we include all control variables.
The remaining of the paper is organized as follows. Section 2 discusses the measurement
of poverty, decent employment and migration and presents the empirical strategy.
Section 3 presents the data and some descriptive statistics. Section 4 discusses the results
and section 5 concludes.
Methodology
This section discusses first the measurement of poverty, decent employment and
migration and secondly presents our empirical strategy.
1. Measurement of the main variables

Poverty
4
Our main indicator of poverty is the annual household consumption per adult equivalent.
This indicator can be a good proxy of household revenues, and in developing countries is
even considered as a better indicator of poverty. As so well stated by Ravallion (1992),
due to large variability of revenues particularly in rural areas and many measurement
errors like recall bias “current consumption will almost certainly be a better indicator
than current income of current standard of living” (Page 13). The adult equivalent allows
to take into account the size of the household, the age of its members and the economies
of scale in consumption. This indicator is directly taken from the computation of the
National Agency of Statistics and Demography (ANSD).
As robustness analysis, we also use a dummy variable indicating if the household is poor
or not. The poverty line is also computed by ANSD and stands for 879 FCFA (1.34 euro)
for urban Dakar, 713 FCFA (1.09 euro) for the other towns and 478 FCFA (0.73 euro) for
rural area.

Decent work
The concept of decent work is since the early 2000s in the heart of the International Labor
Organization (ILO) activities. The ILO (1999) describes decent work as “opportunities for
women and men to obtain decent and productive work in conditions of freedom, equity,
security and human dignity”. Anker et al. (2002) highlight ten aspects to characterize a
decent work: “Employment opportunities”, “Unacceptable works”, “Adequate earnings
and productive work”, “Decent hours”, “Stability and security of work”, “Combining work
and family life”, “Fair treatment in employment”, “Safe work environment”, “Social
protection”, “Social dialogue and workplace relations”.
According to data availability, we try to compute a simple indicator of decent work, easily
understandable and which best approximates the ILO descriptions.
We consider that a worker has a decent work if his/her work meets the following three
criteria:
i.
The worker should not be underemployed
ii.
The worker should be affiliated to a social security system
iii.
The worker should have a stable and secured employment
5
Underemployment is considered according to the definition of the National Agency of
Statistics and Demography (ANSD). A worker is classified as underemployed if he/she
work less than 40 hours per week whereas he/she may wish to work more. This concept
is very important to better understand the functioning of the labor market in developing
countries characterized by low unemployment rates and precarious jobs.
The affiliation to a social security system reflects the social protection aspect which is an
important feature of decent work. Social security system in our data refers typically to
pension fund or health care mutual.
A worker is considered to have a stable and secured employment if he/she has a
permanent contract. Permanent contract indeed reflects a sense of stability since it limits
frequent job changes and unstable labor incomes. We are aware however that permanent
contract is not the only mean to have a stable job.
Robustness tests are performed on the decent work indicator in order to check if there is
not a criterion which alone captures a significant variation of the impact on poverty.

Migration
We aim to consider only economic migration. To qualify a migrant as economic migrant
we take into account two aspects: the age of the migrant and the reason of migration.

The migrant should be aged between 20 and 60 and should migrate for other
reasons than health or study

To not account for students, we just include migrants aged between 25 and 60
among those who have migrated to pursue their study.
2. Empirical strategy
To capture the impact of access to decent employment and migration we estimate the
following equation:
𝑌𝑖 = 𝛼 + 𝛽𝐷𝑖 + 𝛾𝑀𝑖 + 𝛿𝑋𝑖 + 𝜀𝑖
(1)
Index i represent a household.

𝑌𝑖 is an indicator of poverty. In most specifications it represents the
logarithm of the annual consumption of the household per adult equivalent.
In robustness analysis, we replace it by a binary variable taking 1 if the
household is poor and 0 otherwise. In the second part of the empirical
6
analysis, 𝑌𝑖 is the proportion of children aged between 6 and 16 enrolled in
school.

𝐷𝑖 is a binary variable equal to 1 if the household has at least one person
employed in a decent job and 0 otherwise. We describe in the previous
subsection what we mean by decent job. Considering a dummy variable is
relevant since very few households (nearly 0.6%) have more than one
person employed in a decent job.

𝑀𝑖 is a binary variable equal to 1 if the household has at least one migrant
and 0 otherwise. Similarly only 0.7% of households have more than one
migrant which led us to use a binary variable.

𝑋𝑖 is a vector of control variables at the household level susceptible to
explain the household income. 𝑋𝑖 contains: the residence area (urban vs
rural), the number of children (less than 15), the number of adults (more
than 15), the age and the sex of the household head, received transfers from
non-migrants, the proportion of literate adult household members, the
proportion of adult household members with a primary school degree, the
proportion of adult household members with a middle school degree and
the proportion of adult household members with a secondary school degree
and finally dummies for ethnic groups and counties.

𝜀𝑖 is an error term
We estimate (1) using Ordinary Least Squares or Linear Probability Model if 𝑌𝑖 is the
dummy poor or not.
The estimation of 𝛽 and 𝛾 is likely to be biased and non-convergent since 𝜀𝑖 certainly
contains non observables susceptible to be correlated with both the interest variables 𝐷𝑖
and 𝑀𝑖 and the dependent variable 𝑌𝑖 .
Thus we have 𝑐𝑜𝑣(𝜀𝑖 , 𝐷𝑖 ) ≠ 0 and c𝑜𝑣(𝜀𝑖 , 𝑀𝑖 ) ≠ 0.
It is possible that households with persons employed in a decent job or with migrants are
selected into some unobservable characteristics that may be motivation, ability, and
personal relationships etc. creating and endogeneity bias in the estimation of 𝛽 and 𝛾.
Endogeneity can also be due to the reverse causality between decent employment and
poverty, and between migration and poverty. Indeed, poorer households are less likely to
7
meet the necessary conditions to access decent jobs. And similarly, as migration has high
costs, poorer households have more difficulties to finance the migration of one of their
household members.
We use an instrumental variable strategy to deal with these endogeneity problems.
Proxies of network in labor market and network in migration are used as instruments.
Lacking information about the true relationships between people, we approximate social
network using geographical proximity and ethnic group.
The instrument for 𝐷𝑖 is the proportion of households in the county and in the ethnic
group which have at least a member employed in a decent job.
Similarly, the instrument for 𝑀𝑖 is the proportion of households in the county and in the
ethnic group which have at least one migrant.
We name county the second administrative subdivision of Senegal (“département” in
French) after the region. There are 45 counties in Senegal and the survey is representative
at the county level. The number of households by county in our sample varies from 63 to
162. We can imagine that a given household is more likely to have a decent job if a high
number of households in the counties have access to a decent job. The same reasoning
applies to migration. So the instruments are very likely to be correlated with 𝐷𝑖 and 𝑀𝑖 .
Several studies in labor market emphasize the role of network in finding jobs
(Montgomery 1991, Calvó-Armengol and Jackson 2004). Some of them use spatial
interactions to capture the influence of these networks (Wahba and Zenou 2005, Bayer
and al. 2008, Patacchini and Zenou 2012). In migration also, many papers demonstrate
the usefulness of network to overcome migration costs (Munshi 2003, Bertoli 2010,
McKenzie and Rapoport 2010, Beine and al. 2011).
To ensure the exogeneity of the instruments it is necessary to control for county dummies
otherwise a high number of households with decent jobs or a high number of households
with migrants can simply denote the fact that the given county is richer or has
unobservable characteristics correlated to poverty. In this case the instrument is directly
correlated with the household income. To control for county dummies, we should find
instruments that vary inside the county otherwise the two-step procedure will be
impossible. That is why we include the ethnic dimension in defining the instruments.
Instead of taking the proportion of households with migrants in the county as
8
instrumental variable, we take the proportion of households with migrants in the county
by ethnic group. This ensures that the instrumental variable has an intra-county variation.
Likewise the instrument for 𝐷𝑖 would be the proportion of households with decent worker
in the county and by ethnic group.
Ethnicity is relevant because the pattern of migration and access to decent job differs by
ethnic group. Some ethnic groups migrate much more than others. This is also true for the
access in decent jobs but to a lesser extent (see appendix 1). In robustness checks, we use
another instrument for access to decent job using the combination of counties and
employment sectors instead of ethnicity.
For the exclusion restriction to be valid, it is necessary to control for fixed effects of
counties and ethnic groups.
Results
Results are presented in this section. In the first sub-section we analyze the effects of
having a decent job and migration on poverty and different indicators of the household’s
standard of living. In a second sub-section we study the impact of schooling and migration
on children education.
Foremost we assess the validity of the two instruments used to identify a causal effect of
having access to a decent job and migration on poverty. Table 1 report results of a linear
probability model for the first stage regressions with all control variables and fixed effects
for ethnic groups and counties. Results show that the two instruments are strong and
exclusive. Indeed the proportion of household in the county with the same ethnic group
having at least one person employed in a decent job affects significantly, at the one percent
level, the probability of having a decent job in the household and does not affect the
probability to have a migrant. Similarly, the proportion of household in the county with
the same ethnic group having at least one migrant impacts significantly the probability to
have a migrant but is not related to the probability to have a decent worker. This picture
shows that our instruments are good predictors of “decent households” and “migrant
households” and therefore the identification strategy does not suffer from the weak
instrument problem.
9
In addition the Angrist-Pischke test of weak identification is reported in all estimation
results and conclude that instruments are not weak.
Table 1: First stage regressions with all control variables and fixed effects
(1)
decent job
0.877***
(0.0412)
(2)
migrant
-0.0133
(0.0443)
Network migrant
-0.0313
(0.0501)
0.971***
(0.0539)
Ethnic group dummies
County dummies
No. of Observations
R-Squared
F
Yes
Yes
5873
0.211
22.53***
Yes
Yes
5873
0.107
10.11***
Network decent job
Standard errors in parentheses
* p<0.1, ** p<0.05, *** p<0.01
1. Impact on poverty
In the first set of results we compare the effect of decent job and migration on the annual
household’s log consumption per adult equivalent (table 2). We discuss below the
relevancy to use this indicator as measure of poverty.
In all specifications, the impact of decent job and migration is positive and significant at
the 1% level and the elasticity of decent job is always higher except in column 4 where
the difference is not statistically significant. OLS estimation in column 1 with only the two
interest variables indicates an elasticity of 69% for having a decent worker in the
household and 21.2% for migration. The inclusion of additional regressors in column2
halves the elasticity for decent job and increases slightly the elasticity of migration. We
implement a two-stage least squares estimation in columns 3 to 5 instrumenting decent
job and migration. Both elasticities increase in the IV estimates. Column 5 displays our
preferred specification with all control variables and dummies for ethnic group and
counties. This set of dummies ensures that the exclusion restriction of the instruments
holds. Having access to a decent work increases the household consumption by 91.5%
whereas migration increases consumption by 23.1%. The test of difference between these
two elasticities points out that they are different at the 1% significance level. So being
employed in a decent job has higher effect on household consumption than migration.
Regarding the control variables, living in rural area or in the other towns (out the capital)
denotes surprisingly positive effects on consumption compared with living in the capital
(urban Dakar). This effect is confounded by the inclusion of county dummies which are
10
clearly highly correlated with the residence area. As shown in the other specifications
dwelling in rural areas or other towns has negative effects on consumption. The size of
the household has negative effect on consumption. One more child (less than 15) in the
household reduces household consumption by 3.2%. The effect of an additional adult is
twofold, it reduces consumption by 6.3%. The age of the household head and received
transfers from other households inside the country have no significant effect on
consumption. Households headed by women seem to be richer than those headed by men.
Education plays an important role for the household well-being. A percentage point
increase in the proportion of literate adult members in the household increase
consumption by 26.2%. The non-significance of the other education variables seems to be
due to the collinearity with literacy.
1.1.
Impact on the index of living standard
The analysis conducted above may suffer from a limitation that is the consumption of the
household per adult equivalent does not take into account the consumption of people who
migrate since data on consumption are only recorded for local household members. This
view is in line with the idea from Clemens and Pritchett (2008) which suggests to consider
the income of migrants in the measurement of the income per capita of a given country.
The non-record of the migrants’ consumption may lead to an underestimation of the true
effect of migration on the poverty of the household of origin.
To overcome this problem, we replace household consumption by some indicators that
reflect the standard of living of the household using Multiple Correspondence Analysis
(MCA). The indicators are constructed from dwelling characteristics and assets owned by
the households. These characteristics are very likely to capture involvement of migrants
in the household’s standard of living as well as involvement of a decent worker present in
the household. Indeed migrants, even if they do not reside in the household, may certainly
invest in long run assets if they have the capacity to do so.
We construct three different indicators: an indicator of the housing characteristics,
another for the owned assets and the last one is the aggregation of the two first indicators
and reflect general standard of living of the household.
For the construction of the housing and the assets indicators, we firstly choose variables
that discriminate poor and rich households. Variables that concern less than 1% of
11
households are excluded from the analysis. Secondly we run a preliminary MCA and
eliminate variables with smaller contributions in the formation of the first axis.
Table 2: Impact on log total household consumption per capita
Decent job
Migrant
(1)
OLS
log
consumption
0.690***
(0.0395)
(2)
OLS
log
consumption
0.348***
(0.0308)
(3)
IV
log
consumption
2.423***
(0.172)
(4)
IV
log
consumption
1.031***
(0.140)
(5)
IV
log
consumption
0.915***
(0.138)
0.212***
(0.0374)
0.249***
(0.0280)
0.513***
(0.141)
0.853***
(0.113)
0.231**
(0.117)
Zone (ref=Urban Dakar)
Other Cites
-0.447***
(0.0246)
-0.377***
(0.0324)
0.636***
(0.113)
Rural
-0.765***
(0.0270)
-0.665***
(0.0365)
0.345***
(0.113)
# children
-0.0350***
(0.00273)
-0.0353***
(0.00296)
-0.0316***
(0.00271)
# adults
-0.0557***
(0.00282)
-0.0631***
(0.00327)
-0.0634***
(0.00297)
Age of Household Head
-0.00116**
(0.000560)
-0.00115*
(0.000592)
-0.000515
(0.000535)
Household Head female
0.0966***
(0.0175)
0.0936***
(0.0206)
0.101***
(0.0185)
Log Internal Transfers
0.00461***
(0.00143)
0.00673***
(0.00155)
0.000817
(0.00151)
% literate
0.314***
(0.0340)
0.252***
(0.0364)
0.262***
(0.0341)
% primary school degree
-0.0977**
(0.0453)
-0.121**
(0.0480)
0.0420
(0.0443)
% middle school degree
0.0661
(0.0561)
-0.109
(0.0689)
0.0757
(0.0652)
% high school degree
0.179*
(0.0972)
-0.114
(0.129)
0.0770
(0.124)
Constant
Ethnic group dummies
County dummies
No. of Observations
R-Squared
chi2 statistics
F_statistics
p-value test stable=migrant
Angrist-Pischke F-test for weak
instruments (network decent job)
Angrist-Pischke F-test for weak
instruments (network migrant)
12.55***
(0.00963)
No
No
5873
0.0595
169.4***
0.0000
-
13.42***
(0.0416)
No
No
5873
0.398
303.1***
0.0174
-
12.43***
(0.0156)
No
No
5873
.
213.9***
0.0000
390.20***
13.34***
(0.0480)
No
No
5873
0.311
3037.4***
0.3170
275.70***
12.69***
(0.129)
Yes
Yes
5873
0.458
5059.6***
72.45***
0.0001
263.09***
-
-
264.50***
248.62***
182.31***
Robust standard errors in parentheses
* p<0.1, ** p<0.05, *** p<0.01
12
Finally, selected variables for the housing index are: type of housing, roof material, wall
material, flooring material, water source, light source, type of toilet, access to internet and
access to private TV channels.
For computing the asset index, we use the following assets: ventilator, table, chair,
wardrobe, bookcase, living room, phone, computer and fridge.
The standard of living index is built with all the variables used for the housing and assets
index.
In table 3, these index are considered as dependent variables in columns 1 to 3. Column 4
reports our preferred specification with all controls in column 5 of table 2 for comparison
purpose. We run an IV estimates in all columns with all control variables and fixed effects.
Access to a decent job has a strong positive and significant effect for all the three indexes
with a smaller elasticity for the housing index and higher effect for the assets index.
Migration has no effect in the housing index but impact significantly the ownership of
assets. Compared to the impact on the consumption, decent job and migration seem to
have higher effect on the standard of living index with respective elasticities of 92.8% and
35.9%.
Results for control variables are almost the same than for previous results on
consumption except that we have a positive effect of internal transfers on the assets
ownership and the general standard of living and positive effect of all the education
variables for the three indexes.
For the three indexes, the impact of access to decent job is higher and statistically different
from the impact of migration, suggesting that decent employment is more efficient for
fighting against poverty.
13
Table 3: Impact on household's standard of living indicator
(4)
IV
log consumption
1.099***
(0.173)
(3)
IV
Standard
of living
0.928***
(0.132)
0.168
(0.126)
0.610***
(0.171)
0.359***
(0.137)
0.231**
(0.117)
Other Cites
0.822***
(0.0826)
0.891***
(0.139)
0.888***
(0.0982)
0.636***
(0.113)
Rural
0.0556
(0.0812)
0.304**
(0.138)
0.148
(0.0971)
0.345***
(0.113)
# children
0.00223
(0.00317)
0.00302
(0.00394)
0.00256
(0.00327)
-0.0316***
(0.00271)
# adults
0.0156***
(0.00344)
0.0425***
(0.00421)
0.0283***
(0.00355)
-0.0634***
(0.00297)
Age of Household Head
0.000297
(0.000620)
-0.000627
(0.000730)
-0.0000704
(0.000622)
-0.000515
(0.000535)
Household Head female
0.203***
(0.0204)
0.126***
(0.0276)
0.185***
(0.0221)
0.101***
(0.0185)
Log Internal Transfers
0.000528
(0.00165)
0.00650***
(0.00208)
0.00298*
(0.00172)
0.000817
(0.00151)
% literate
0.402***
(0.0358)
0.454***
(0.0455)
0.448***
(0.0371)
0.262***
(0.0341)
% primary school degree
0.269***
(0.0448)
0.232***
(0.0623)
0.275***
(0.0487)
0.0420
(0.0443)
% middle school degree
0.411***
(0.0590)
0.566***
(0.0902)
0.514***
(0.0669)
0.0757
(0.0652)
% high school degree
0.394***
(0.114)
0.489***
(0.169)
0.466***
(0.126)
0.0770
(0.124)
Constant
-0.583***
(0.113)
5855
0.636
235.7
0.0021
263.09***
-0.962***
(0.168)
5850
0.447
92.53
0.0384
263.67***
-0.767***
(0.127)
5864
0.617
212.3
0.0021
263.02***
12.69***
(0.129)
5873
0.458
72.45***
0.0001
263.09***
182.04***
181.76***
182.19***
182.31***
decent job
migrant
(1)
IV
Housing
(2)
IV
Assets
0.698***
(0.125)
0.915***
(0.138)
Zone (ref=Urban Dakar)
No. of Observations
R-Squared
F
p-value test stable=migrant
Angrist-Pischke F-test for weak
instruments (network decent job)
Angrist-Pischke F-test for weak
instruments (network migrant)
Robust standard errors in parentheses
* p<0.1, ** p<0.05, *** p<0.01
1.2.
Robustness checks
We implement a number of robustness checks in table 4 to test if our main result always
hold. All results come from IV estimates with all the control variables.
14
In column 1, we use a dummy variable that takes 1 if household is poor and 0 if it is not
poor. As explained previously, this measure may be more suitable in the analysis of
poverty. Results in column 1 show that access to decent job reduces significantly poverty
at the 1% level whereas the coefficient for migration is negative but not significant. This
strongly confirms our main result that decent employment does better than migration in
reducing poverty.
In columns 2 to 4 we run some robustness analysis in the measure of decent employment.
We use three alternative measures and for each we eliminate one of the three criteria
keeping only two criteria. “Decent1” keeps the underemployment and the social security
system criteria; “decent2” considers underemployment and having a permanent contract;
“decent3” are for those who are affiliated in a social security system and have a permanent
contract.
For the three alternative measures, results show significant positive impact of decent job
in the consumption per adult equivalent and always greater than the impact of migration
and the difference is still significant. The elasticities are smaller for “decent1” and
“decent2”, the magnitude of “decent3” is almost similar to that of our baseline result.
In column 5, we use another instrument for access to a decent job. One might argue that
ethnic group is not a good proxy of network in labor market since we do not have a clear
pattern of differences in the access to decent employment between ethnic groups
contrary to migration. So we use as instrument a proxy of network that exploits the
geographical proximity as previously and the employment sector of the household head.
Clearly, instead of considering the proportion of households with at least one person
employed in a decent job in the county and in the ethnic group as previously, we consider
as instrument the proportion of household with decent worker in the county and in the
employment sector of the household head.
The different sectors are: public
administration, public company, big private and non-financial company, small private and
non-financial company, bank or insurance company, international organization, embassy
or consulate, associative company or individual business. People in the same employment
sector may be more likely to help to find a job than people in the same ethnic group. This
new instrument is used in column 5 for access to decent job. For ensuring that the
exclusion restriction always hold, it is necessary to control for dummies of the
15
employment sector that we do not show in the table because of space constraints2. Results
show positive and significant effect of decent employment and migration and still a higher
elasticity for decent employment but not statistically different from that of migration. This
relatively low elasticity of decent employment is due to the fact dummies for sectors
account for much of the variability of decent employment that explains household
consumption.
In column 6 we remove the employment sectors dummies and the elasticity of decent
employment rises from 32.1% to 59.3% and this time statistically different from the
elasticity of migration.
We restrict in column 7 migration to developed countries. In fact nearly 40% of migrants
in our sample live in other African countries which are developing countries. This pattern
tends to lower the net impact of migration on the living standards of native households.
So we restrict migration to European countries, the US and Canada what accounts for 60%
of all migrant households and 4% of all households. Results show that the elasticity of
decent employment remains almost the same as our baseline model but the elasticity of
migration rises from 23.1% to 36.0%. Thus, migration to developed countries increases
the impact of migration on household consumption but the impact of decent employment
is still higher and statistically different.
2
Available upon request
16
Table 4: Heterogeneity and robustness checks
(1)
poverty
decent job
migrant
(2)
log
consumption
(3)
log
consumption
(4)
log
consumption
(5)
log
consumption
0.321***
(0.0933)
(6)
log
consumption
0.593***
(0.0614)
0.235*
(0.124)
0.268**
(0.115)
0.267**
(0.117)
0.215*
(0.113)
0.210*
(0.114)
-0.348***
(0.0821)
-0.0933
(0.0941)
decent1
(7)
log
consumption
0.909***
(0.138)
0.659***
(0.108)
decent2
0.773***
(0.119)
decent3
0.921***
(0.118)
Migrant developed
0.360**
(0.178)
Other Cites
-0.129**
(0.0529)
0.610***
(0.109)
0.611***
(0.113)
0.613***
(0.111)
0.521***
(0.106)
0.582***
(0.106)
0.631***
(0.113)
Rural
-0.158***
(0.0523)
0.310***
(0.109)
0.327***
(0.113)
0.342***
(0.111)
0.221**
(0.105)
0.281***
(0.106)
0.343***
(0.114)
# children
0.0189***
(0.00222)
-0.0316***
(0.00273)
-0.0315***
(0.00268)
-0.0320***
(0.00274)
-0.0321***
(0.00260)
-0.0319***
(0.00264)
-0.0317***
(0.00272)
# adults
0.0338***
(0.00234)
-0.0661***
(0.00306)
-0.0639***
(0.00295)
-0.0640***
(0.00299)
-0.0601***
(0.00280)
-0.0620***
(0.00282)
-0.0638***
(0.00302)
Age of Household
Head
-0.000385
-0.000818
-0.000544
-0.000226
-0.000465
-0.000669
-0.000507
(0.000407)
(0.000533)
(0.000535)
(0.000542)
(0.000542)
(0.000522)
(0.000535)
-0.0521***
0.0981***
0.0957***
0.109***
0.0959***
0.0888***
0.0990***
(0.0138)
(0.0182)
(0.0183)
(0.0186)
(0.0173)
(0.0173)
(0.0188)
0.00456***
(0.00108)
0.00154
0.00123
0.00118
0.000876
0.000690
0.00117
(0.00151)
(0.00150)
(0.00151)
(0.00146)
(0.00148)
(0.00155)
% literate
-0.103***
(0.0227)
0.259***
(0.0349)
0.271***
(0.0334)
0.224***
(0.0351)
0.288***
(0.0323)
0.291***
(0.0328)
0.260***
(0.0342)
% primary school
degree
-0.0618**
0.0397
0.0248
0.0743*
0.0488
0.0418
0.0376
(0.0283)
(0.0438)
(0.0442)
(0.0438)
(0.0422)
(0.0432)
(0.0447)
-0.0749**
0.114*
0.0954
0.000827
0.169***
0.149***
0.0731
(0.0381)
(0.0620)
(0.0634)
(0.0685)
(0.0537)
(0.0552)
(0.0654)
-0.0207
0.191*
0.103
-0.186
0.223**
0.211**
0.0815
(0.0673)
(0.112)
(0.118)
(0.135)
(0.0935)
(0.101)
(0.124)
0.133**
(0.0677)
5873
0.217
30.25***
0.0398
12.75***
(0.123)
5873
0.464
73.75***
0.0063
12.70***
(0.128)
5873
0.468
74.39***
0.0015
12.68***
(0.128)
5873
0.457
71.91***
0.0000
12.98***
(0.133)
5873
0.499
75.09***
0.4773
12.76***
(0.120)
5873
0.485
79.17***
0.0029
12.70***
(0.128)
5873
0.458
72.89***
0.0138
263.09***
374.41***
339.84***
339.79***
368.05***
513.78***
263.48***
182.31***
183.00***
178.52***
180.51***
185.10***
184.56***
60.64***
Household Head
female
Log Internal
Transfers
% middle school
degree
% high school
degree
Constant
No. of Observations
R-Squared
F
p-value test
decent=migrant
F-test for weak
instruments
(network decent)
F-test for weak
instruments
(network migrant)
Standard errors in parentheses: * p<0.1, ** p<0.05, *** p<0.01
17
2. Impact on children’s education
Results in the previous sub-section have shown that access to decent job has a greater
impact on poverty than migration. One of the main channels is that a decent job promotes
stability in the household and allows household members to look to the future, to invest
more in economic activities and in children’s human capital. We compare in this section
the effect of decent employment and migration on children’s enrolment in school. Our
dependent variable is now the proportion of children aged between 6 and 16 enrolled in
school in the household. Results are presented in table 5.
In column 1 and 2 we run simple OLS estimation and see that the effect of decent
employment is significantly higher than that of migration. In column 3 the introduction of
education variables greatly reduces the magnitude of the coefficient of decent
employment which is still higher than that of migration but not statistically different. In
column 4 to 8 we instrument access to decent job and migration. The effect of decent job
remains significant in all specifications except in column 6 but the effect of migration is
no longer significant. In columns 5 and 6, we use our main instrument that is network
based on county and ethnic group both for decent job and migration. In column 6, the
effect of decent employment disappears when we introduce county dummies. This is
probably due to the fact that children’s schooling is highly correlated to geographical
factors which annihilate the effect of access to decent employment. In columns 7 and 8,
we modify the instrument for decent job using the network based on counties and
employment sectors like in columns 5 and 6 in table 4 and we get a significant effect of
decent job. In column 8, we introduce the log consumption of the household and the effect
of decent job remains significant at the 5% level. This is a powerful result that indicates
all things being equal, taking two households with the same level of consumption, one
with a person employed in a decent job and the other with no decent worker, the
household with decent worker will invest more in children’s schooling.
Results for control variables show that households headed by elderly or headed by a
woman, invest more in children's education. All control variables related to the education
of adult household members are positive and highly significant indicating that the
education of adult household members is one of the main drivers of investment in
children’s education.
18
Table 5: Impact on children schooling
decent job
migrant
(1)
OLS
%
schooled
children
0.284***
(0.0156)
(2)
OLS
% schooled
children
(3)
OLS
% schooled
children
(5)
IV
% schooled
children
(6)
IV
% schooled
children
(7)
IV
% schooled
children
(8)
IV
% schooled
children
0.0685***
(0.0179)
(4)
IV
%
schooled
children
0.811***
(0.0837)
0.195***
(0.0168)
0.210**
(0.0819)
-0.0350
(0.0794)
0.118***
(0.0371)
0.0827**
(0.0375)
0.0982***
(0.0200)
0.0677***
(0.0198)
0.0443**
(0.0188)
0.0846
(0.0702)
0.0723
(0.0655)
0.00208
(0.0740)
0.0129
(0.0731)
-0.00208
(0.0740)
log consumption
0.0731***
(0.0109)
Zone (ref=
Urban Dakar)
Other Cites
0.00318
(0.0192)
-0.00572
(0.0186)
0.00852
(0.0205)
0.0775
(0.0762)
0.103
(0.0748)
0.0845
(0.0747)
-0.217***
(0.0204)
-0.135***
(0.0196)
-0.116***
(0.0227)
-0.0724
(0.0761)
-0.0419
(0.0742)
-0.0383
(0.0740)
-0.00674***
(0.00192)
0.00227
(0.00179)
0.00229
(0.00180)
0.00231
(0.00172)
0.00239
(0.00172)
0.00411**
(0.00172)
-0.00168
(0.00194)
-0.00829***
(0.00182)
-0.00918***
(0.00196)
-0.00646***
(0.00185)
-0.00725***
(0.00180)
-0.00308
(0.00196)
0.000997**
0.00124***
0.00128***
0.000739**
0.000809**
0.000844**
(0.000413)
(0.000385)
(0.000388)
(0.000376)
(0.000374)
(0.000373)
0.0500***
0.0244**
0.0291**
0.0219*
0.0282**
0.0218*
(0.0128)
(0.0120)
(0.0127)
(0.0126)
(0.0124)
(0.0123)
-0.000357
-0.000739
-0.000678
0.00124
0.00126
0.000990
(0.000985)
(0.000915)
(0.000941)
(0.000974)
(0.000974)
(0.000968)
% literate
0.0826***
(0.0237)
0.0736***
(0.0241)
0.0884***
(0.0241)
0.0781***
(0.0236)
0.0647***
(0.0234)
% primary
school degree
0.519***
0.515***
0.416***
0.414***
0.403***
(0.0294)
(0.0294)
(0.0297)
(0.0295)
(0.0292)
0.480***
0.449***
0.409***
0.375***
0.355***
(0.0329)
(0.0365)
(0.0365)
(0.0332)
(0.0331)
0.462***
0.366***
0.475***
0.372***
0.353***
(0.0636)
(0.0839)
(0.0854)
(0.0681)
(0.0676)
Rural
# children
# adults
Age of Household Head
Household Head
female
Log Internal
Transfers
% middle school
degree
% high school
degree
Constant
Ethnic group
dummies
County
dummies
No.
Observations
R-Squared
F
p-value test
stable=migrant
F-test for weak
instrument
(network
decent)
F-test for weak
instruments
(migrant)
0.599***
(0.00607)
No
0.696***
(0.0298)
No
0.496***
(0.0306)
No
0.571***
(0.00891)
No
0.480***
(0.0322)
No
0.450***
(0.0870)
Yes
0.419***
(0.0851)
Yes
-0.526***
(0.159)
Yes
No
No
No
No
No
Yes
Yes
Yes
4870
4870
4870
4870
4870
4870
4870
4870
0.0309
175.3***
0.0000
0.128
98.35***
0.0000
0.251
157.7***
0.3439
47.25***
-
0.245
150.2***
-
0.321
43.45***
-
0.324
44.78***
-
0.332
45.53***
-
-
-
-
232.16***
146.51***
128.13***
363.15***
355.03***
-
-
-
259.69***
241.24***
171.78***
175.19***
174.35***
19
Conclusion
This paper analyzes the impact of decent employment and migration on poverty and
schooling for Senegalese households. I capture decent work through variables related to
the stability of work and social protection and use different poverty index to conduct this
study. Proxies of networks of migration and decent employment are used as instruments
to control self-selection bias. Results show that migration reduces poverty by 23.1%
confirming the existing literature. But the impact of access to decent employment is four
times higher than that of migration and 2.5 times higher if we restrict migration to
developed countries.
One possible explanation is that decent employment allows people to have a forwardlooking behavior, to think more about their future and to invest more in their family
members’ human capital and well-being in general.
References
Acosta P., C. Calderon, P. Fajnzylber and H. Lopez (2008): "What is the Impact of International
Remittances on Poverty and Inequality in Latin America?" World Development, 36(1), 89-114.
ANSD (2013) : Situation Economique et Sociale du Sénégal en 2011, version définitive. 344 p.
ANSD (2016) : Situation Economique et Sociale du Sénégal en 2013, version définitive. 350 p.
Barham B. and S. Boucher (1998): "Migration, remittances, and inequality: estimating the net
effects of migration on income distribution" Journal of Development Economics, 55(2), 307-331.
Bayer P., Ross S. L. and Topa G. (2008) : "Place of work and place of residence: informal hiring
networks and labor market outcomes". Journal of Political Economy. 116, 1150–1196.
Bertoli S. (2010): "Networks, sorting and self-selection of Ecuadorian migrants," Annales
d'Economie et de Statistique, 97/98, pp. 261-288.
Camara N. F. et Gueye A. S. (2013) : "Déterminants de l’accès à l’emploi au Sénégal", Mémoire de
fin de formation ITS. ENSAE (Sénégal).
Cazes S. and Verick S. (2013): Perspectives on labour economics for development. International
Labour Office, Geneva. 305 p.
Chiwuzulum Odozi J., Taiwo Awoyemi T., and Titus Omonona B. (2010): “Household Poverty and
Inequality: The Implication of Migrants’ Remittances in Nigeria”, Journal of Economic Policy
Reform 13, 191–199.
Clemens M. A. and Pritchett L. (2008): Income per Natural: Measuring Development for People
Rather Than Places. Population and Development Review. Vol. 34, Issue 3, pp. 395-434.
20
Gubert F., Lassourd T., and Mesplé-Somps S. (2010) : "Transferts de fonds des migrants, pauvreté
et inégalités au Mali. Analyse à partir de trois scenarii contrefactuels", Revue Economique 61,
1023–1050.
Gupta S., Patillo C. & Wagh S. (2009) : “Impact of remittances on poverty and financial
development in Sub-Saharan Africa”. World Development, vol. 37(1), pp. 104-115.
Gutierrez C., Orecchia C., Paci P. and Serneels P. (2007) : "Does employment generation really
matter for poverty reduction ? ", World Bank, Policy Research Working Paper N° 4432.
Imai K. S., Gaiha R., Ali A. and Kaicker N. (2012): "Remittances, growth and poverty. New evidence
from Asian countries", discussion papers produced by the Asia and the Pacific Division, IFAD N°
15. 40 p.
International Labor Organization (1999): Decent Work: Report of the Director General,
International Labour Conference, 87th Session.
Lachaud J. P. (1999) Envois de fonds, inégalité et pauvreté au Burkina Faso, Revue Tiers Monde,
Programme National Persée 40(160), 793–827.
Margolis D. N., Miotti L., Mouhoud E. M. and Oudinet J. (2015) : ""To Have and Have Not":
International Migration, Poverty and Inequality in Algeria", Scandinavian Journal of Economics.
117(2), 650–685.
Mbaye L. M. (2014): "Barcelona or Die": Understanding Illegal Migration from Senegal", IZA
Journal of Migration, 3:21.
Mishra P. (2007): "Emigration and wages in source countries: Evidence from Mexico," Journal of
Development Economics, 82(1), 180-199.
Montgomery J. (1991): "Social networks and labor-market outcomes: toward an economic
analysis". American Economic Review. 81, 1408–1418.
Munshi, K. (2003) : "Networks in the modern economy: Mexican migrants in the U.S. labor
market". The Quarterly Journal of Economics. 118 (2), 549–599.
Patacchini E. and Zenou Y. (2012), "Ethnic networks and employment outcomes", Regional Science
and Urban Economics, 42. pp. 938-949.
Ravallion M. (1992): "Poverty comparaisons. A guide to concepts and methods", Washington,
Lsms, Working papers n°88, World Bank.
World Bank (2015): Migration and Development Brief 24, World Bank, Washington, DC.
21
Appendix
Percentage of migrants and decent workers by ethnic group
Ethnic groups
wolof/lébou
sérère
poular
soninké
diola
mandingue/socé
balante
bambara
malinké
mandiack/mankagne
maure
others
% households with migrant
6.01
3.69
7.44
15.31
3.84
7.26
1.28
6.87
7.94
4.29
10.61
7.32
22
% households with decent worker
7.20
6.99
4.47
9.18
9.04
4.95
3.85
5.34
6.35
11.43
4.55
2.44
First stage regression
(1)
Stable job
0.877***
(0.0412)
(2)
migrant
-0.0133
(0.0443)
-0.0313
(0.0501)
0.971***
(0.0539)
Other Cites
-0.154***
(0.0374)
-0.0401
(0.0402)
Rural
-0.175***
(0.0368)
-0.0584
(0.0396)
# children
-0.000682
(0.00107)
0.00183
(0.00115)
# adults
0.00433***
(0.00112)
0.00522***
(0.00121)
Age of Household Head
-0.000489**
(0.000210)
0.000592***
(0.000226)
Household Head female
-0.0390***
(0.00691)
0.0485***
(0.00743)
Log Internal Transfers
-0.000210
(0.000572)
-0.00242***
(0.000616)
% literate
0.0720***
(0.0119)
0.0192
(0.0129)
% primary school degree
0.00870
(0.0162)
0.0402**
(0.0175)
% middle school degree
0.228***
(0.0200)
0.0361*
(0.0216)
% high school degree
0.359***
(0.0367)
0.0108
(0.0395)
Constant
0.142***
(0.0447)
Yes
Yes
5873
0.211
22.53***
-0.0247
(0.0481)
Yes
Yes
5873
0.107
10.11***
Network stable job
Network migrant
Zone (ref=Urban Dakar)
Ethnic group dummies
County dummies
No. of Observations
R-Squared
F
23
.2
.1
% of poor
.2
.1
0
0
no decent job
decent job
no migrant
migrant
.6
.4
0
.2
.2
.4
.6
.8
% of enrolled children
1
.8
Proportion of schooled children by household with decent job and migrant household
0
% of poor
.3
.3
Poverty rates by household with decent job and migrant household
no decent job
no migrant
decent job
migrant
Descriptive statistics
consumption
% enrolled
children
% decent job
% migrant
urban Dakar
other urban
areas
rural area
n_enfant
n_adulte
agecm
genre_cm
ltransfert
alphab_men
cfee_men
bfem_men
bac_men
Observations
5873
Mean
390374.3
Standard
deviation
356965.1
4870
5873
5873
5873
.6205568
.061638
.0631704
0.0905
.3951458
-
5873
5873
5873
5873
5873
5873
5873
5873
5873
5873
5873
0.4240
0.4855
4.0819
5.20756
51.63545
1.251149
202449.2
.4802441
.1523479
.078649
.0174824
3.388492
3.269512
14.49304
.4337112
508881.1
.3275504
.2100838
.1648137
.0806209
24
Minimum
11998.16
Maximum
5628477
0
1
0
1
17
1
0
0
0
0
0
43
28
99
2
9425000
1
1
1
1
25