DEPARTMENT OF ECONOMICS
UNIVERSITY OF MILAN - BICOCCA
WORKING PAPER SERIES
Xenophobic Attacks, Migration Intentions
and Networks Evidence from the South of
Africa
Guido Friebel, Juan Miguel Gallego and Mariapia
Mendola
No. 213 – November 2011
Dipartimento di Economia Politica
Università degli Studi di Milano - Bicocca
http://dipeco.economia.unimib.it
1
Xenophobic Attacks, Migration Intentions and Networks:
Evidence from the South of Africa
1
Guido Friebel - Goethe University Frankfurt, IZA and CEPR
Juan Miguel Gallego – Universidad del Rosario and LdA
Mariapia Mendola – University of Milan Bicocca and LdA
September 2011
Abstract
We investigate how emigration flows from a developing region are affected by xenophobic violence
at destination. Our empirical analysis is based on a unique survey among more than 1000
households collected in Mozambique in summe 2008, a few months after a series of xenophobic
attacks in South Africa killed dozens and displaced thousands of immigrants from neighbouring
countries. We estimate migration intentions of Mozambicans before and after the attacks,
controlling for the characteristics of households and previous migration behaviour. Using a placebo
period, we show that other things equal, the migration intention of household heads decreases from
37 to 33 percent. The sensitivity of migration intentions to violence is larger for household heads
with many children younger than 15 years, decreasing the migration intention by 11 percentage
points. Most importantly, the sensitivity of migration intentions is highest for those household heads
with many young children whose families have no access to social networks. For these household
heads, the intention falls by 15 percentage points. Social networks provide insurance against the
consequences young children suffer in case the household head would be harmed by xenophobic
violence and consequently could not provide for the family.
Keywords: violence, risk, migration, household behaviour, Mozambique
JEL: O1, R2, J6, D1
1
We thank Alice Mesnard, Melissa Myambo, Luca Stanca, Pedro Vicente and participants at the NORFACE-CREAM
Conference on ‘Migration, Economic Change and Social Challenge’ at UCL in London, the CEPR-TOM Conference on
'International Migration: Transnational links, Effects and Policies' at Venice International University in Venice, the IZA
Workshop on ‘Legal and Illicit Immigration: Theory, Empirics and Policy’ in Bonn and a seminar at the Goethe
University in Frankfurt for comments and discussion. Financial support from Centro Studi Luca d'Agliano (LdA) for
data collection is gratefully acknowledged. All errors are ours. Contact information: gfriebel@wiwi.uni-frankfurt.de,
juan.gallego@urosario.edu.co, mariapia.mendola@unimib.it (corresponding author).
2
Migration is one of the most important ways through which workers can improve their productivity
and wages and increase their families’ welfare. While the overall effects of migration on host and
source countries alike seem to be beneficial, there have been through history fears of natives in host
countries that migrants increase the unemployment among domestic workers and lower wages.
History is full of examples in which domestic workers took violent actions against immigrants in
order to deter migration. Examples include the anti-immigrant violence in 19
th
century New York,
which was documented by Asbury (1927), the recent violence against Hispanic migrants in the US,
or the xenophobic violence in Eastern Germany after German re-unification.
It remains an open question how this form of violence affects migration and migration
intentions. We here investigate how the migration flows from a developing region are affected by
xenophobia and violent actions in a more developed host country. We investigate a hand-collected
survey among more than 1000 households, administrated in Mozambique in summer 2008, a few
months after a series of xenophobic attacks in South Africa killed dozens and displaced tens of
thousands of migrants from neighbouring countries. Our identification strategy is based on the
comparison of migration intentions of Mozambicans before and after the attacks occurred in May
2008, and on the use of a placebo period in which no violence occurred. We control for differing
characteristics of the household samples (from a wide range of survey demographic measurements)
and previous migration behaviour. We find that other things equal, the intention of the head of the
household to migrate after the attacks is lower than before; it decreases from 37 percent to 33
percent. The sensitivity of migration intentions to violence is larger for household heads with many
children younger than 15 years, decreasing the migration intention by 11 percentage points (p.p
thereafter). Evidently, people are not only concerned about their own health, but also about the
welfare of their offspring. Most importantly, the sensitivity of migration intentions is highest for
households with many young children and little access to social networks. For such households, the
intention falls by almost 15 p.p. Social networks thus provide insurance against the consequences
small children suffer when the household head is harmed by xenophobic violence and consequently
cannot provide.
The small but growing body of literature on violence and migration has focused on out-
migration decisions in high-level violence environments (see Mesnard 2009 on out-migration flows
from Colombia, for example). Yet, very little is known on how violence and xenophobic feelings in
host countries affect migration intentions and behaviour in source countries. To the best of our
knowledge this paper is the first that measures how violence or other types of xenophobic behaviour
in host countries affect migration intentions, and to what extent social networks in the source
country may provide insurance against the risks associated with migration.
3
Our findings relate to intentions to migrate. A host of papers has looked at migration
intentions (for instance, Burda et al., 1998; Drinkwater and Ingram, 2008; Epstein and Gang, 2006;
Fouarge and Ester, 2007; Lam, 2000; Liebig and Sousa-Poza, 2004; Papapanagos and Sanfey, 2001;
and Ubelmesser, 2006). Manski (1990) has questioned the general use of migration intentions data
as a proxy for actual migration, but emigration intentions have been shown to be a good predictor of
future actual emigration behaviour (see, for example, van Dalen and Henkens, 2008). It has also
been argued that intentions are a monotonic function of the underlying driving variables that
motivate migration (Burda et al.1998). Finally, using migration intentions data avoids the sample
selection difficulties that arise from the use of the host country data (see Liebig and Sousa-Poza,
2004; van Dale and Henkens, 2008).
Our results shows that violence in the host country reduces the intention to migrate.
Violence has the smallest effect on people who live in small households and have few children, and
the largest on people with many household members whose family is badly connected to social
networks. Migration intentions of household heads with large families who have a good social
network are much less sensitive to violence than the ones with bad networks. This confirms that
social networks are an important insurance mechanism in developing economies. However, the joint
facts that (i) better access to social networks reduces the sensitivity of migration intentions to
violence although (ii) access to networks in Mozambique cannot provide protection against violence
in South Africa, lend itself to the interpretation that migrant workers seem to care more about the
future of their offspring than about their own health.
The next section provides the necessary background on the two economies, Mozambique
and South Africa, on the xenophobic violence in 2008, and the role of family and networks in
Mozambique. Section 2 presents an illustrative model, and Section 3 the data. Section 4 discusses
our empirical strategy that builds on the difference between the before and after violence intention
to migrate, and the comparison with a placebo period that cannot be affected by the violence.
Section 5 presents the main results, and Section 6 discusses a number of potential concerns about
statistic identification. We then conclude.
1. Migration from Mozambique, violence, and the role of family and networks
Mozambique is one of the poorest countries in the world with a GDP per capita of $446 at current
US dollars in 2008. Out of 22.7 millions population, 70 percent of Mozambicans live below the
poverty line- and 35 percent below the extreme poverty line (PRSP, 2007). Mozambique has had
slow economic growth until the beginning of the 1990s, poor levels of education, especially of
4
women, low productivity and a lack of employment opportunities. Basic infrastructure is lacking in
many rural areas, whence the isolation of communities and poor integration of rural-urban markets
(AfDB/OECD, 2003 and 2008). Faced with such poverty, migration is one of the few ways to
improve the situation of a family. Because of its geographic proximity and much more advanced
development, South Africa (SA) is the main destination of Mozambican migrant workers. SA is the
African superpower and ranks as an upper-middle income economy with a GDP per capita of
$5.666 in current US dollars in 2008. The economic gap between Mozambique and SA in terms of
development and growth is striking. Hence, South Africa seems to offer large economic advantages
to potential migrants in Mozambique.
TABLE 1 ABOUT HERE
Panel A of Table 1 shows recent statistics about the stock of international migrants from
Mozambique to OECD countries collected by Docquier et al. (2009). This dataset is collected from
a population census in host countries where information about (legal) migration based on country of
birth is available. The figures are quite similar to our representative household survey from two
southern Mozambican Provinces shown in Panel B: the bulk of migrant workers from Mozambique
go to South Africa.
Our empirical strategy is based on the comparison of migration intentions before and after
xenophobic attacks, that is, the third quarter in 2007, compared to the third quarter in 2008, and the
placebo period, the first quarter in 2008. The next table indicates that macroeconomic changes are
not a driver of changes in intentions that we will observe.
TABLE 2 ABOUT HERE
Table 2 presents figures on inflation and GDP variation. There is no particular change in the
third quarter (our period of reference) in 2007 and 2008 in both countries. The inflation rate in 2007
and 2008 in both countries follows the same trends in all terms of the year. At the end of 2007
cumulative price variation was 8.2% in Mozambique and 7.1% in South Africa. The numbers are
slightly higher in 2008 (10.3% and 11.5% respectively). GDP has the same trend, without large
changes between quarters in the two years, with the exception of the third quarter of 2008 where
GDP in Mozambique fell by 0.14%. If anything, this fall in the GDP should have triggered stronger
migration intentions, had the xenophobic attacks not occurred.
