Migration, Remittances and Household Welfare in Ethiopia Lisa Andersson, Department of ...

34 Pages·2012·0.71 MB·English

by  Lisa Andersson

Migration, Remittances and Household Welfare in Ethiopia

Lisa Andersson, Department of Economics
University of Gothenburg
[email protected]
This version May, 2012
Work in progress
___________________________________________________________________________

Abstract
This paper investigates the effects of international remittances on
household welfare in rural Ethiopia. Using a recently conducted
migration survey with rich information on migration and remittance
patterns, a matching technique is applied in order to address self-
selection bias. By using information before and after the households
started receiving remittances this study sheds light on the change in
welfare that households experience after starting receiving
remittances. The results show that remittances have a significant
impact on a welfare variable that has previously not received much
attention in the migration literature, namely household subjective
wellbeing. The effect on consumer- and productive asset accumulation
is more ambiguous.

Keywords: remittances, migration, Ethiopia, propensity score matching

1
1. Introduction

According to official World Bank statistics, about 30 million Africans have migrated
internationally. Demographic factors are likely to increase the African migration rates
substantially over the next decades (World Bank 2011a). Millions of households in the
migration sending countries are already affected by migration through remittances sent back
to the migrants‟ origin countries. Remittance inflows to the continent have seen a fourfold
increase in the past 20 years, and were estimated at nearly 40 billion USD, 2.5 % of GDP, in
2010. The inflow of remittances to Africa exceeds the amount of official aid and is the second
largest source of net foreign capital inflow after foreign direct investments (FDI) (World
Bank, 2011a).
Consequently, the economic impact of migration and remittances has received an increasing
interest from both researchers and policy makers. Remittances are often the most
straightforward link between migration and welfare of the households in the origin country.
According to the new economics of labor migration (NELM), migration is part of a household
strategy to overcome market failures such as imperfect credit and insurance markets.
Remittances that provide households with an income not correlated with farm income can
loosen production and investment constraints and finance investments in new production
technologies and input. Apart from the direct effects of remittances, there can also be
multiplier effects on income, employment and production in the migration sending country
(Taylor 1999). When emerged in the 1980‟s and 1990‟s, NELM contrasted a previously
dominating but more pessimistic view on migration and development that argued that
remittances often are used for non-productive investments and lead to the development of
passive, non-productive and remittance-dependent communities (de Haas, 2007).
This paper investigates the impact of migration and remittances on household welfare in the
origin country by looking at household asset accumulation and subjective wellbeing in rural
Ethiopia. Despite the increase in migration and remittance flows to the continent, literature on
international migration and development in Sub-Saharan Africa is relatively scarce, largely
due to data constraints (Lucas, 2006). Ethiopia is an interesting country to study in this
context since it is one of the top 10 remittance receiving countries in Sub-Saharan Africa, and
the inflow of remittances to the country has increased dramatically in the past years, from 46
million USD in 2003 to an estimated number of 389 million UDS in 2010 (World Bank
2011a). This study makes use of a new and rich data set on migration and remittances from
2
Ethiopia in order to estimate the impacts of migration and remittances on household welfare.
Two measures are used to define welfare - household assets and household subjective
wellbeing. Household subjective wellbeing reflects the household‟s own rating of its
economic situation and captures the direct impact of remittances on household welfare, while
measures of asset accumulation can increase our understanding of the specific impacts
remittances might have on household asset investments. For this purpose two asset indices,
one for consumer assets and one for productive assets, are constructed.
Previous empirical findings from cross-country analysis show remittances to have a positive
impact on poverty (Adams and Page, 2005 (71 developing countries) and Acosta et al, 2008
(Latin America)). When it comes to spending and investment patterns, Adams and Cuecuecha
(2010a) find that households in Guatemala who receive remittances spend more on the margin
on two investment goods; education and housing, while the same authors find that remittance-
receiving households in Indonesia spend more on the margin on one key consumption good -
food, but on the other hand decreases their marginal spending on a key investment good;
housing (Adams and Cuecuecha 2010b). A few studies have specifically investigated the link
between remittances and asset accumulation. Adams (1998) investigates the impact of internal
and international remittances on asset accumulation in rural Pakistan. The study finds a
positive and significant relationship between remittances and two types of physical assets:
irrigated- and rainfed land. Quisumbing and McNiven (2010) assess the impact of internal
migration and remittances on assets in rural Philippines using longitudinal data and an
instrumental variable approach. The study finds that having large number of migrant children
in the household reduces the values of nonland assets. However, remittances have a positive
impact on housing, consumer durables and nonland assets.
Subjective wellbeing has received very little attention in previous studies related to the effects
of remittances on the welfare of households left behind. We are only aware of one previous
study, by Semyonov and Gorodzeisky (2006), which uses a subjective measure of wellbeing
to investigate the link between remittances and household welfare in the origin country. In
this study the authors derive their measure of subjective wellbeing from a combination of two
measures; the households‟ own evaluation of its capability in meeting its everyday basic
needs and its self-assessed relative position compared to the average Filipino family. The
study finds that most of the difference in standard of living between households with overseas
workers and those without are attributed to remittances.
3
A challenge when estimating the causal impact of migration and remittances on household
welfare is self-selection. If migration is not a random decision, and remittances are not
randomly assigned, there might be confounding factors that influence both the probability of
migration/receiving remittances and the outcome of interest, which would result in biased
estimates of the impact of remittances on the outcome. In this study, a matching approach will
be applied in order to address the possible self-selection issue. Treatment will in this case be
whether the household receives remittances (or have a migrant in the additional analyses), in
order to measure the average treatment effect of the treated. The advantage with this approach
is that it allows us to compare households that receive remittances with otherwise similar
households that do not receive remittances in order to mitigate the self-selection bias. Since
the data contains retrospective information about household assets and subjective wellbeing
five years ago, as well as information about when the household started receiving remittances,
we are able to look at the change in welfare before and after households start receiving
remittances.
The reminder of the papers is structured as follows: section 2 describes the migration and
remittance patterns in Ethiopia; section 3 gives an overview of the data and descriptive
statistics. Section 4 describes the methodology used. The results and robustness checks are
presented in section 5 and section 6 concludes.
2. Migration and remittance patterns in Ethiopia

The character, direction and volume of international migration flows from Ethiopia have gone
through a number of changes in the past decades. Three factors can be identified as the main
drivers of the country‟s migration patterns in the past three decades: political instability,
decline or stagnation in the agricultural sector, and the 1980s government resettlement
program. Revolution and an unstable political climate in Ethiopia shaped the migration flows
during the 70‟s. Most of the people who migrated at this time belonged to a well-educated
urban part of the population and migrated to western countries in order to seek political
asylum political migration was followed by more economically oriented migration, initially
driven by the aspirations of the urban population. Today, as the Middle East has become an
important destination region for Ethiopian migrants, the migrants are to an increasing extent
from rural areas migrating to find better (employment) opportunities abroad (Geda and Irving,
4
2011). According to the World Bank‟s Migration and Remittances Factbook (2011), the top
emigration destination countries for Ethiopians are Sudan, the United States, Israel, Djibouti,
Kenya, Saudi Arabia, Canada, Germany, Italy, and Sweden (World Bank, 2011b).
The World Bank ranks Ethiopia to be the 8th largest remittance receiver in Sub-Saharan Africa
in 2010, with an inflow of remittances reaching 387 million USD, to be compared with net
Foreign Direct Investment inflows of 100 million USD and net Overseas Development
Assistance (ODI) at 3.3 billion USD (World Bank, 2011b). The numbers used by the World
Bank rely on the International Monetary Fund‟s (IMF) Balance of Payments statistics. There
is however a large discrepancy between the numbers recorded by the IMF and officially
recorded remittance inflows reported by the National Bank of Ethiopia. In particular, the
National Bank reports remittances inflows of about $600 million while the actual volume of
remittances, when taking into account flows through both formal and informal channels, are
estimated to be in the range of $1 billion to $2 billion annually (Geda and Irving, 2011).
Despite the large and increasing inflows of remittances, very little is known about the impact
that these remittances might have on the households and the country‟s economy as a whole.
Using descriptive analysis, Mohapatra et al. (2009) show that households that depend on
international remittances face fewer shocks from food shortages and drought compared to
other households. Remittance-dependent households also reported fewer shocks measured as
illness of household members. The authors hypothesize that better nutrition, leading to better
health, as one possible explanation to this link, although the difference compared to other
households were smaller than the differences observed for food shortages. However, given the
descriptive nature of this study it is difficult to draw conclusions about a causal link between
remittances and household welfare.
3. Data and Descriptive statistics

The data used in this study comes from the newly collected IS Academy: A World in Motion
Migration and Development household survey, funded by the Dutch Ministry of Foreign
Affairs and administered by the Maastricht Graduate School of Governance. A sample of
1,280 randomly selected households was interviewed between March and May 2011. The
sample includes households with current migrants abroad, households with migrants who
returned from abroad, and households with no international migration experience. The
definition of a household in this survey follows a definition previously used in other migration
5
surveys where the concept of a household is extended to not only include members who are
„living together and have communal arrangements concerning subsistence and other
necessities of life‟, but also those members who presently resides elsewhere (in the country or
abroad) but whose „principle commitments and obligations are to that household‟ (see e.g.,
Ünalan, 2005). A person living abroad can in this way still be considered a household
member.
The survey was administered across five different regions throughout the country: Amhara,
Oromia, Southern Nations Nationalities and People‟s Region (SNNPR), Tigray, as well as the
capital Addis Ababa, which together account for 96% of the country population. In each
region, three different Woredas (districts) were selected for sampling, totalling 15 data
collection sites in both urban and rural areas. The sampling followed a two-stage sampling
procedure in order to ensure that enough migration households were captured in the survey.
First a listing was conducted at each site to identify households as a migrant-, return-, or non-
migrant household. Based on this identification, households were randomly selected for
enumeration in each site, ensuring that a satisfying level of households with migration
experience was included in the survey. A migrant was in the survey defined as a person who
is living in another country and has been away for at least three consecutive months. The
questionnaire includes detailed questions about the migration and remittance experiences of
the household. In addition, questions about education, assets, income, expenditures, and
subjective wealth of the households are included.
Because the main focus of this paper is on the effects of migration and remittances on
household welfare in the rural areas of Ethiopia, the sample will be restricted to only include
rural households. There are in total 780 rural households in the data. An advantage with the
data we have at hand is that we can separate migration and remittance patterns of the
households, while many other datasets only include information on either migration or
remittances. The data also contains information about previous experience of international
migration through a member who migrated but has now returned to the household. Among
the non-remittance receiving households, 94 households have a return migrant. Households
with a return migrant in the household, and who possibly also received remittances in the
past, might differ from the other households in the sample and in order to avoid any bias of
the results we exclude the return migrant households from the sample. There are also a
number of households in the sample with only one member. Since by definition a household
with only one member would be excluded from the survey if the only member migrated (and
6
leave no one behind to be interviewed) we also exclude single households from the analysis.
Furthermore, the data also contains information regarding the point in time when the
household started receiving remittances. A large majority of the remittance receiving
households, about 92%, started receiving remittances in the past five years, which indicates
that migration and remittances is a quite recent phenomenon in the rural areas of Ethiopia. By
combining this information with retrospective information about asset holdings today and five
years ago we can better control for the true impact of remittances on assets and we therefore
exclude the households that sent a migrant or started receiving remittances more than five
years ago. After seven additional households were dropped due to missing values in the
outcome variables, the final sample used in the analysis consist of 651 households. Out of
these households, 30% (197 households) have at least one member abroad, and 19% (125
households) receive remittances.1 The large majority of the remittances senders are members
of the remittance receiving households.2
3.1. Descriptive statistics
The migrants in the sample reside in different parts of the world. The most common migration
destination countries in our sample are Saudi Arabia (24%), USA (20%), Sudan and United
Arab Emirate (12%) and South Africa (8%). Other destinations include Israel, Qatar, Kuwait,
Canada, UK and Yemen. The survey records how many times the households have received
remittances in the past 12 months. About 7% of the households receives remittances every
second months or more frequently, 12.5% received remittances every third month and 17%
received remittances two times. Most respondents hence state that the remittances were
irregular (27%) or that they received remittances once (33%) during the past 12 months. The
total value of the remittances received during the past 12 months also varies quite a lot, from
500 Birr to 110,000 Birr, with a mean value of 11727 Birr.3 The households are also asked to
rate how important remittances are for the household‟s total income. The data contains some
missing values for this variable, but of those households who answered about 10% state that
international remittances accounts for 100% of their income, and 30% of the households state
that remittances from abroad account for 50% of their income or more. Most households

1 By remittances we here refer to monetary remittances. The data also contains information about remittances in-
kind, but since monetary remittances are far more common and in-kind remittances often complements the monetary
remittances (only two households in the sample receive in-kind remittances without receiving monetary remittances)
we restrict the analyses to only look at monetary remittances.
2 There are however 18 households who only receive remittances from non-members of the household. The analyses
are performed both including and excluding these households, and the results remain very similar.
3 The average monthly income in the sample is 1,915 Birr (which corresponds to a yearly income of 22,980 Birr), but
it should be noted that the income variable contains some missing values. Note: 1 Birr≈0.057USD.
7
(39%) state that they mainly spend the remittances received on daily needs such as food, debt
repayment (19%) and housing/land (13%) are the second and third most common ways to
spend the remittances, followed by ceremonies and agriculture (5%), durable goods and
education (4%). 4
Table 1 presents descriptive statistics for all rural household included in the sample and
separately for migrant-, remittance- and non-migrant households respectively.
[Table 1 about here]
As mentioned before, 30% of the households have a migrant abroad and 19% of the
households receive remittances. We also see that about 54% of the migrant households
receive remittances. There is hence quite a significant part of the migrant households who do
not receive remittances. Since the analysis in this paper is restricted to only look at
households who started receiving remittance in the past five years, the fact that only about
half the migrants sends remittances may partly be explained by the amount of time that the
migrants have been away. Among the migrants, 37% have been away between 3 and 12
months if we look at all the migrants in the sample. However, the share of migrants who
migrated between 3 and 12 months ago is much higher among the non-remittance sending
migrants, at almost 57%.
When we compare the overall sample with households by migration- and remittance status we
find that migrant- and remittance receiving households are on average larger, have more
members in working age, a higher female to male ratio and a higher education level compared
to the control group (i.e., households without migrants and remittances in the last column of
table 1). Migrant- and remittance receiving households are less likely to have a household
head who is employed or have a business, but more likely to have a head involved in
agricultural activities. Households who receives remittances and/or have a migrant also more
often have children below the age of 18 years old compared to the control group, although
migrant households without remittances have less likelihood of having children in the
households compared to both the migrant households with remittances and households with
no migration or remittances. The variables and their likely impact on the probability of
receiving remittances will be discussed in more detail in section 5.1.

4 Looking directly at how the remittances are spent is interesting and can give useful insights, but might not
necessarily tell us much about the impact of remittances on household expenditures and investments since, as pointed
out by Taylor (1999), money is fungible. Spending remittances on daily needs will free up resources that can be spent
on other things or invested in productive activities.
8
4. Methodology

In this paper household welfare will be measured using both a measure of the household‟s
subjective wellbeing and using an asset index strategy. In order to address the problem of self-
selection a propensity score matching approach will be used. Although the data origins from a
cross-section dataset, the survey contains retrospective questions regarding asset holdings and
subjective wellbeing of the household five years ago. By taking advantage of the fact that
most households started receiving remittances during the past five years, outcomes can be
measured in terms of changes in assets and subjective wellbeing before and after the
households started receiving remittances. This enables the use of a „difference-in-difference‟
approach to more precisely measure the effects of remittances.
4.1 Propensity score matching
One of the main challenges when estimating the causal effects of remittances on welfare
measures is self-selection. There might be unobservable characteristics that affect both the
probability that the household receives remittances (and migrates) and the outcome of interest.
If selection into sending remittances is not a random decision, analysis of the effect of
remittances on household welfare will give biased estimates unless the problem of self-
selection is addressed.
Previous studies has used a number of approaches to address selectivity into migration and
remittance sending, including assuming selection on observables (e.g. Adams, 1998),
parametric selection correction models (e.g. Barham and Bucher, 1998), instrumental
variables (e.g. Mansuri, 2006; McKenzie and Rapoport, 2007), and propensity score matching
(Esquivel and Huerta-Pineda, 2007; Cox-Edwards et .al, 2009). In this paper the last method
is applied.
Propensity score matching is often used in a program evaluation setting, where the objective
is to compare participant outcome with and without treatment. The method was first proposed
as a way to reduce bias in estimation of treatment effects with observational data in the
seminal work by Rosenbaum and Rubin (1983), and has become a popular method to measure
the impact of economic policy interventions (Becker and Ichino, 2002). The idea is to first
create an index that summarizes observable characteristics of the households into a propensity
score index. The households are then divided into two groups, those who receive remittances
and those who don‟t, and ranked according to their propensity score. Finally the households
9
are matched with similar households from the other group. In this way households in the
treatment group can be matched and compared with households from the control group who
have similar characteristics in every aspect except that they don‟t receive remittances.
In equation form, our goal is to estimate the causal treatment effect:
(1)

Where and is the outcome with and without treatment respectively for household .

Consider * + to be a binary indicator where 1 equals being assigned into treatment and
0 means not being assigned treatment. The Average Treatment Effect (ATE) can be estimated
through:
, - , - (2)

ATE is hence the average difference between the treated households (in this context treated
households are households who receives remittances) and the non-treated households.
However, such comparison might not capture the true impact of the treatment if we have
selection into treatment and there are other factors that are correlated with both treatment and
some omitted variable that is affecting the outcome variable. A fundamental problem is that
we can observe the outcome variable under either treatment or control for each household, but
never both at the same time.
In this context, a preferred parameter to use instead of ATE is the Average Treatment effect on
the Treated (ATT), defined by:
, - , - (3)
Where , - is never observed. Replacing , - by the expected value of
, -, which is observable in ATE, would not give an accurate estimate as long as
for the treated and the comparison group systematically differ. The true parameter is only
identified if:
, - – - (4)
As discussed above, this is not very likely to hold in non-experimental studies. Instead we
rely on a matching approach in order to derive a counterfactual that enables us to match
treated households with non-treated households with as similar characteristics as possible in
order to reduce the bias from self- selection. The matching is made based on an index, the
10