Migration and multidimensional well-being in Ethiopia: investigating the role of migrants destinations

Abstract

The purpose of this paper is to better understand the relationship between migration and multi-dimensional well-being in the context of Ethiopia. We investigate the differences in well-being between migrant, non-migrant and return migrant households. We then go a step further to understand the role of migrants destinations for well-being by disaggregating migration destination to the Middle East, Africa and the North. We find that migrant and return migrant households are better off in terms of well-being than non-migrant households. Furthermore, the findings underline the importance of taking migrants destinations into account in determining the wellbeing of the households left behind. While households with a migrant in the North are significantly more likely to report higher well-being outcomes than non-migrant households, this is not true for households with migrants in other destination regions.

Keywords:

• Migration and development
• return migration
• Ethiopia
• well-being

Introduction

The linkages between migration and development have received increasing attention in recent decades both within policy circles and the academic literature (Skeldon, 2008; de Haas, 2010). There are several connections between migration and development at the household level including financial remittances, social remittances, health, and education, among others. However, the majority of this literature either examines remittance and non-remittance receiver or migrant and non-migrant households. This study takes a novel approach by comparing migrant, non-migrant and return migrant households and including interactions with remittances. Furthermore, frequently studies also only focus on one corridor of interest and do not account for variation in destinations of migrants. Qualitative research has sought to examine the impact of the destination state on remittance sending (Mahmud, 2016), but few studies have investigated the relationship between migrants’ destination and the household left behind, which is done in this paper. This paper is based on a unique data-set that first enables a comparison between migrant, non-migrant and return migrant households and second, allows for an investigation into the different migrant flows that are relatively evenly distributed across three primary migration corridors of the North, Middle East, and the South.

The third core contribution of this paper is the use of a multidimensional well-being index to examine the relationship between migration and the households left behind. Multidimensional well-being indices provide a more holistic approach to understanding poverty by going beyond monetary income to include factors such as education, health, and social inclusion (Alkire, 2002, 2005; Alkire & Foster, 2011). Although the relationship between variables such as migration and education or migration and health has been explored in the literature, few studies have taken a multidimensional approach to migration and overall wellbeing (Adams, 2010, 2013; Lipton, 1980). In this paper we have developed and utilized a multidimensional well-being index that is relevant for our case study of the Ethiopian context. While many papers look at deprivation (poverty indices), this paper looks at being well (well-being index).¹

Under the implementation of the new European Union Partnership Agreement, Ethiopia has become a central country of investment for funding on migration issues. Despite the rapid increase in investment in Ethiopia, little is known regarding the impacts of migration on households left behind, well-being and development in the country. Within the current policy environment in Ethiopia there is a strong focus on ‘migration prevention’, which contradicts several migration theories that migration can increase well-being for households left behind. Within this context, understanding the relationship between migration and well-being of households left behind is important to both inform policy and increase understandings of the unique dynamics occurring in Ethiopia.

This paper has five key sections. The first provides an overview on the multidimensional well-being approach to migration. The second section gives a short background on migration from Ethiopia. The third section details the data and empirical strategy of the paper. The forth presents the results and the final section provides a discussion and conclusion.

The Linkages between Migration and Well-being

In this section we provide a brief overview of four linkages between migration and wellbeing: first, we examine the overall relationship between migration and wellbeing, second between return migration and wellbeing, third we specifically assess the role of remittances, and forth we look at the role of destinations within the migration and wellbeing debate.

Research has demonstrated that it is not the poorest of the poor who migrate (de Haas, 2010, 2012). In order to migrate, individuals need to have some level of resources to be able to finance the costs of the migration and to have access to migration (de Haas, 2010). Carling (2002) terms those that have the aspiration to migration, but lack the ability as involuntary non-migrants. Barriers to mobility include not only economic resources, but also lack of social networks, physical dangers and barriers to migrating irregularly, qualitative constraints such as skill level, and practical constraints (Carling, 2002).

In terms of wellbeing, de Haas’s (2007) work shows that non-migrants are often poorer as they do not have the basic resources to migrate in the first place. As Carling (2002) shows, those who would like to migrate but are not able to may be in a similarly worse off position. It may be precisely the factors that make someone worse off that do not allow the person to migrate. For these reasons, we hypothesise that non-migrant households will be worse off than migrant households (H1).

A key element of the household migration strategy is to be able to send remittances to the household back home. Ample research has demonstrated that remittances increase household wellbeing by lowering household financial constraints and enable them to invest in areas such as human and physical capital (see e.g.: Acosta, Calderon, & Lopez, 2008; Adams & Page, 2005; Ajaero, Tzeadibe, Obisie-Nmehielle, & Ike, 2018; de Haas, 2012; Kapur, 2003).We therefore predict that remittance receiving households will be better off than non-migrant households (H2).

Regarding return migration, there are contrasting notions as to the role of return migration in influencing household well-being upon return. Primarily, this debate stems from the reason for the return migration. From a theoretical perspective, neoclassical economic theory suggests that return migration is the unintended result of a failed migration experience (Cassarino, 2004; de Haas, 2010). This differs from the New Economics of Labour Migration (NELM) theory that views return migration as the final stage of a successful migration cycle wherein the individual returns as their choice because they have achieved their migration goals (Cassarino, 2004; de Haas, 2010). Following from a NELM perspective, we hypothesise that return migrant households will be better off than non-migrant households (H3).

In general, the role of migrant destinations has received little attention within the literature. It is increasingly recognized that little is known about South-South migration trends and more specifically, how they compare to South-North migration trends (Bakewell, 2009; Ratha & Shaw, 2007). The 2013 International Organization for Migration (IOM) World Migration Report highlights the changing nature of migrant destinations from previous flows of primarily South-North migration to increasing flows of South-South and North-South migration (IOM, 2013). Fourtypercentof migrants were cited as migrating from the South to the North, one-third from the South to the South, 22 percent from the North to the North and 5 percentfrom the North to the South (IOM, 2013). One key challenge noted in this report is the accuracy of statistics on South-South migration, which are arguably the most difficult to capture.

We conceptualize the flows in this paper into three categories of South-South, South-North, and South-Middle East due to the fact that some Middle East countries can be classified as south or north depending on the indicators that are used. For example, the World Bank classifies Saudi Arabia as ‘North’ based on income, while the Human Development Index of UNDP classifies Saudi Arabia as ‘South’ (Bakewell, 2009). Resultantly, the wealth of the Middle East attracts large numbers of low-skilled migrants: for women this is primarily in domestic work and for men in construction, the service industry, and also domestic work. However, from a rights and development perspective, evidence demonstrates high numbers of migrant human rights violations in the Middle East. The unique characteristics of migration flows to the Middle East of low skilled labour in a context of low human rights arguably legitimize looking at this flow separately from South-South and South-North migration. Hence, we examine these flows separately, due to the challenges and nature of migration to the Middle East we hypothesise that households with a migrant in the North will be better off than households with a migrant in the South or Middle East (H4).

Although it is well-established that remittances increase household well-being, we assume that migrants in the North have different capabilities for sending remittances than migrants in the South or Middle East and that migrants in the north have better capabilities for sending higher amounts of remittances. That is, migrants in the North are most likely able to send more remittances. We therefore hypothesise that households with a remittance sender in the North are more likely to be better off than households with a remittance sender in the South or Middle East (H5).

A Multidimensional Well-being Approach to Migration

The multi-dimensional approach to poverty and well-being originates from the pioneering work of Amartya Sen and has been expanded upon by many scholars (Laderchi, Saith, & Steward, 2003; Nussbaum, 1992, 2000; Ravallion, 1994; Sen, 1976, 1982, 1985, 1993; Thorbecke, 2008).

The concept of multi-dimensional poverty/well-being has been further operationalized by academics over the years since the concept was first established (Bastos, Fernandes, & Passos, 2004; Baulch & Masset, 2003; Bourguignon & Chakravarty, 2003; Bradshaw & Finch, 2003; Klasen, 2000; Perry, 2002). The underlying idea is that poverty is more than just monetary poverty, but that there can be deprivation in many other areas. Several studies suggest that the use of monetary and multidimensional poverty measures result in different depictions of poverty (Klasen, 2000; Perry, 2002; Baulch & Masset, 2003; Bastos et al., 2004; Whelan, Layte, & Maitre, 2004). Assessing poverty or well-being from a multidimensional perspective allows for addressing other key areas such as health, education, living standards, physical safety as well as income. By using such a measure, the complexity of well-being can be investigated in a more holistic way.

Using a multi-dimensional approach is helpful in understanding overall well-being as only looking at a traditional indicator of income can be misleading and not give the full picture. For instance, in India, income growth has been increasing but child malnutrition has stayed the same (Citizens’ Initiative for the Rights of Children Under Six, 2006). At the same time, according to the Oxford Poverty and Human Development Initiative (OPHI), people themselves often describe their situations as being multi-dimensional. OPHI found that poor people depict poverty as relating to variables such as: poor health, nutrition, lack of adequate sanitation and clean water, social exclusion, low education, bad housing, violence, shame, and disempowerment.

Understanding the different areas or dimensions of well-being also allows for more targeted policy approaches. Indicators always need to be chosen based on the specific country context (Alkire, 2008).Here we make sure that all indicators that are chosen for this paper are specific to the context of Ethiopia. This was done based on our work and knowledge of the country as well as from consultation with local partners in Ethiopia.

Utilising a multidimensional approach to poverty/well-being has recently been applied to the field of migration (Gassmann, Siegel, Vanore, & Waidler, 2012, 2013; Loschmann & Siegel, 2015; Siegel & Waidler, 2012; Vanore & Siegel, 2013). Most of the previous work examining how migration affects well-being outcomes has been focused on Mexico and concentrated on income poverty, education, and health (Kanaiaupuni & Donato, 1999; Kandel, 2003; McKenzie, 2005; McKenzie & Hildebrandt, 2005; Mckenzie & Rapoport, 2007; McKenzie & Rapoport, 2011). Until recently, the relationship between migration and different development outcomes such as income, expenditure, education and health have all been looked at separately and rarely a more holistic approach has been taken (Adams, 2010, 2013; Adams & Page, 2005; Lipton, 1980). This paper utilizes a holistic approach to wellbeing to better understand the relationship between migration and well-being.

Migration from Ethiopia: An Overview of Current Trends

Migration from Ethiopia has been increasing since the end of the conflict period in 1991 and particularly over the past decade. The recent migration streams from Ethiopia are primarily for labour migration as opposed to the previous flows that were primarily characterized by refugee migration (Kuschminder & Siegel, 2014). Despite experiencing high levels of growth in the past decade, Ethiopia is one of the poorest countries in the world; ranking 173 out of 187 countries measured by the UNDP Human Development Index (2015). Ethiopia has high levels of unemployment, particularly among urban youth. Urban unemployment has a strong gendered dimension with 11.4 percent of urban males being unemploymend compared to 24.2 percent of urban females in 2012 (Bilgili, Kuschminder, and Siegel, 2017). Following from this, emigration from Ethiopia is highly gendered with 60 percent of current migrants being female (Kuschminder, Andersson, & Siegel, 2012).

At present, there are three central migration destination regions from Ethiopia. The first and most prominent is to the Middle East. There is an increasing body of research on female emigration from Ethiopia primarily to the Middle East (de Regt, 2010; Fernandez, 2010, 2013; ILO, 2011; Kuschminder, 2016, 2017; RMMS, 2014). This research includes drivers for migration, conditions abroad, experiences of return and reintegration, and concerns regarding the human rights and wellbeing of the migrants. Most these migration flows for women are for domestic work. The primary destination country for men to the Middle East is Saudi Arabia, where male migrants mainly work in the construction sector. Yemen has been a key transit country in route to Saudi Arabia and there is also a growing body of research highlighting the concerns of Ethiopian migrants’ wellbeing in Yemen, particularly given the increasing violence in Yemen in 2015 (see RMMS, 2014).

The second key destination is migration within Africa, wherein the primary destination of Ethiopian migrants is South Africa. Primarily young men migrate to South Africa for economic purposes (Horwood, 2009). To some migrants South Africa is the final destination country, but South Africa is also used as a transit country for migration further afield, such as to the US, Europe or Canada. Approximately 4,000 Ethiopian migrants are also apprehended in Tanzania each year en route to South Africa. The number of apprehensions in other countries along this route such as Kenya, Uganda and Mozambique is not known.

The third key destination of migrants from Ethiopia is to North America and Europe. There are several important differences with this destination as compared to Africa and the Middle East. Migrants going to the North are more likely to be educated, migrate legally with documents, and have mixed reasons for migration such as study purposes or family reunification (Kuschminder et al., 2012). Irregular migration from Ethiopia to the North does also occur, although precise figures on this are unknown.

The majority of Ethiopian migration is currently for labour purposes; however, over 61,000 Ethiopians had pending asylum cases in 2014 (United Nations High Commissioner on Refugees, 2015). The majority of these claims were lodged in non-industrialized countries, predominantly South Africa, Kenya and Yemen, with only 5,262 asylum claims from Ethiopians lodged in industrialized states in 2014 (UNHCR population statistics, 2014).

Finally, return migration has also become an increasingly salient issue in Ethiopia. In October 2013, Saudi Arabia removed over 150,000 Ethiopians back to Addis Ababa, which has become known as the ‘Ethiopian return crises’. This created a humanitarian emergency as shelter and basic needs were required for many of these returnees. Return to Ethiopia, however, construes many different types of migrants. This includes: professional and transnational returnees, labour returnees, student returnees, domestic worker returnees from the Middle East, Assisted Voluntary Returnees from Europe, Egypt or Yemen, and forced returnees.

At this time, there is little evidence as to the impact of migration on household wellbeing in Ethiopia. In a recent study based on the same data used in this paper, Andersson (2014) finds that rural remittance receiving households are more likely to have positive perceptions of their subjective well-being and the position of their household compared to others in the community. Rural remittance receiving households were also found to be able to accumulate more consumer assets that non-remittance receiving households. This paper continues to build on these findings by examining multidimensional wellbeing from migration in general, not just remittance receiving households.

Empirical strategy

This study employs a multidimensional poverty approach to analyse the link between migration and well-being. The methodology builds on the multidimensional poverty methodology developed by Alkire and Foster (2011) that extends traditional poverty measures to include several dimensions but uses the cut-off method of Roelen and Gassmann (2012) and Roelen, Gassmann, and de Neubourg (2012) for simplicity. The identification involves two forms of cut-off. The first cut-off concerns deprivation within a specific dimension, by considering several indicators related to the dimension. The second cut-off relates to an aggregated measure of the overall deprivation of the household, taking all the different dimensions included in the analysis into account.

A household is considered to be well-off in a given indicator if the threshold set for the given indicator is met. Indicators and cut offs were chosen based on data availability as well as the authors’ knowledge of the Ethiopian context, consultation will other experts and what has been used previously in the Ethiopian context (Ambel, Mehta, & Yigezu, 2015). Appendix 1 provides an overview of each dimension, indicator, thresholds and descriptive statistics. For example, the indicator for electricity within the housing dimension will take on value 1 if the household meets the corresponding threshold of having access to electricity, and value 0 if the household does not have access. Electricity is seen as important for well-being in Ethiopia and the government has been partnering with donors to increase access (World Bank, 2016). All indicators within each dimension are evaluated against their thresholds, and used in order to establish well-being rates for each dimension. A household is considered well in a given dimension if it meets the corresponding threshold with respect to each indicator.

The choice of cut-off levels is normative, and depends on the set of indicators and dimensions included (Alkire & Foster, 2011). In this paper, we use a cut-off level of 0.7 and assign equal weights to all indicators in each dimension. This means that in a case where a dimension has two indicators, the household needs to be well-off in both dimensions to meet the requirement. In the case of four indicators in one dimension, the household needs to be well-off in three out of four indicators to be considered well-off in that dimension.

After establishing the thresholds for each dimension, an overall wellbeing index is created by aggregating the different dimensions. Again, the cut-off is set to 0.7, and all dimensions are given equal weight. This means that a household needs to be well-off in 70 percent of the dimensions in order to be considered well-off in the overall multidimensional poverty measure. A variable taking on value 1 if the household meets this requirement and the overall well-being is above 70 percent when all dimensions have been aggregated, and value 0 if the well-being is below 70 percent, is created.

In this study, five different dimensions of well-being are used to measure multidimensional poverty. The dimensions are: education; housing; health; income, employment and assets; and social inclusion. Each dimension includes between two and four indicators. The education dimension includes two indicators: household head is literate; children of school age (7–15) in the household goes to school. Since the school attendance indicator only applies to households with children in school age, households without children in school age are only evaluated based on the literacy of the household head. The housing dimension includes the following four indicators: access to electricity; access to a toilet; the house has more than one room; the house has appropriate flooring (not dirt, sand or dung). The health dimension includes four dimensions: access to a health clinic; household is able to meet its food needs; household does not have any disabled or seriously ill household head; household has not lost a child. The indicator regarding the ability of meeting the food needs is based on a subjective question where the household is asked to assess its ability to meet its food needs. The income, employment and asset dimension is also based on four indicators: household is earning the threshold of 2 dollars a day per adult equivalent; none of the children in the household are working; the household has more than one income source; the household owns at least two consumption goods. Income per adult equivalent is calculated following the OECD adult equivalence factor. The first adult in the household is given a weight of 1, and each additional adult member is assigned a weight of 0.7. Children, defined as members in the age of 0–13, are assigned value 0.5.² The income measure of 2 dollars is calculated according to the World Bank conversion factor for private consumption. The fifth dimension, inclusion, includes two indicators: household owns a mobile phone; the household is a member of at least one organization.

Finally, probit regressions are used to estimate the probability that a household is well-off, both when it comes to the aggregated overall wellbeing measure, and in each of the included dimensions. The dependent variable is in the former case a binary variable taking on value 1 if the household is well-off in the aggregate multidimensional indicator, and value 0 otherwise. In the latter case, each dimension is tested separately using a binary variable that takes on value 1 if the household is considered well-off in a given dimension. The variable of interest is a set of migration variables. In a first step, migration is measured though an aggregate measure taking on value 1 if the household has a member that migrated abroad. In a second step, the migration variable is disaggregated into three binary variables to take into account if the migrant member is located in a destination country in the North, the Middle East or within the African continent. A number of control variables on individual and household level are also included in the specification.

Data and descriptive statistics

This paper is based on the IS Academy Migration and Development: A World in Motionproject data collection in Ethiopia. An in-depth household survey was conducted of 1284 households across five different regions in Ethiopia from March to May 2011. Surveys were made with three types of households: households that currently had a member living abroad; households that had a member who had lived abroad and returned; households that had no experience of international migration. It is important to stress that the migrants themselves were not interviewed and all data used in this paper is from interviewing the household in Ethiopia. The surveys were conducted in the following five regions of Ethiopia: Amhara, Oromia, Southern Nations Nationalities and People’s Region (SNNPR), Tigray, and Addis Ababa, which together account for 96 percentof the population. In each region, three different woredas (districts) were selected for sampling, totalling 15 data collection sites. The sampling strategy was based on a two-stage approach. First a listing was made at each site to identify households as migrant, return migrant or non-migrant households. Based on this identification, households were randomly selected for enumeration in each site with an equal proportion of migrant or return migrant households to non-migrant households. The data is not nationally representative and cannot be generalized to represent all Ethiopian migration.

A migrant was defined in this study as any member of the household who had been living in another country for a minimum of three consecutive months. Similarly, a return migrant is defined as any household member that lived abroad for a minimum of three months and had since returned for a minimum of three months. These definitions were chosen to include seasonal migration, which occurs annually for a shorter period, usually three to eight months.

It is possible that a household contains more than one migrant. These multiple migrants may in turn reside in different destination regions. In the sample of 426 migrant households, 15 households had multiple migrants who reside in different geographic areas. Since a goal with the current study is to compare migration across different regions, these 15 households were dropped from the data. The final data-set contains 411 migrant households and 127 return migrant households, which corresponds to 32 and 10% of the household sample.

The data also contains information on remittances received by the households. Remittances were defined as international monetary transfers that were received by a household from a migrant within the past 12 months since the time of interview. There are 278 households (22%) in the sample who receive remittances. A majority of these households, 234 households (84%), receive remittances from members of the household who emigrated abroad. There are however 44 households that receive remittances without having a migrant member. A unique element of this survey was that remittances were captured from any source, meaning that a non-household member could be sending remittances to the household (such as a family friend or relation abroad).

Migrants were also examined further by migrant destination region: North, Middle East and Africa. All migrant destinations except for Russia and Israel are classified into these three destination groups, which represents 97% of all migrants in the sample. Table 1 shows the percentage of migrants per destination country.

Table 1. Migrant household by destination region of migrants.

Destination	Percentage
North	30
Middle East	20
Africa	47
Other	3
Total	100

Note:

North includes USA, Canada, Australia, and all European countries except Russia.

Many migrant households have migrants residing in the Middle East (47%), while about 30% of the households have migrants in the North and 20% in other African countries outside Ethiopia. The most common destination countries are Saudi Arabia, United States, United Arab Emirates and Sudan. Table 2 displays some basic characteristics of the households in the sample, by migrant region.

Table 2. Household characteristics, summary table by migration destination region.

Characteristics	All households	North	Africa	Middle East
Urban	0.39	0.68	0.20	0.35
Gender (1 = female)	0.35	0.54	0.25	0.36
Age of household head	50.25	59.49	50.80	52.32
Number of members 65 and above	0.23	0.55	0.21	0.23
Number of children 0–14	1.55	0.72	1.55	1.61
Head in paid work	0.16	0.11	0.06	0.16
Head self employed	0.16	0.13	0.06	0.11
Head farmer	0.34	0.05	0.64	0.38
Head retired	0.13	0.38	0.05	0.09
No education	0.36	0.07	0.44	0.37
Primary education	0.23	0.13	0.28	0.27
Secondary education	0.33	0.48	0.26	0.30
Graduate education or higher	0.08	0.33	0.02	0.05
Receive remittances	0.22	0.63	0.45	0.58
Total	1268	104	87	222

Note:

All households include the full sample of households, with or without migrants.

The overview shows that households with migrants in the North are predominantly urban, while households with migrants in other African countries or the Middle East are predominantly rural. While a little over half of the heads (54%) in households with a migrant in the North are female, this is much less common in households with migrants in Africa and the Middle East (25 and 36% respectively). Heads in households with migrants in the North also tend to be more educated and older compared to those with migrants in other regions. The share of households receiving remittances also varies depending on in which region the households have a migrant. Among households with migrants in the North, 63% receive remittances.

The corresponding number of households with migrants in the Middle East and Africa that receive remittances is 58 and 45% respectively.

A previous study, using the same data-set, has also shown differences in the characteristics of Ethiopian migrants depending on their region of destination. Migrants in the North are generally more educated and migrate for multiple reasons (including family reunification, professional and education opportunities and security), while migrants in the Middle East and Africa are less educated and migrate mainly for economic reasons to find employment (Kuschminder et al., 2012).

Results

Table 3 shows descriptive statistics for the five dimensions of the multidimensional well-being index as well as the aggregated multidimensional indicator and its association with migration.

Table 3. Multidimensional poverty indicators, summary table by migration status and migration destination region.

Dimension	Overall	Migrant household	Return migrant household	Non-migrant household	North	Africa	Middle East
Education	0.5	0.49	0.54	0.5	0.67	0.45	0.41
Health	0.9	0.9	0.9	0.89	0.99	0.8	0.9
Housing	0.39	0.46	0.47	0.35	0.88	0.21	0.33
Income/assets	0.25	0.29	0.39	0.21	0.51	0.18	0.24
Inclusion	0.42	0.5	0.47	0.36	0.68	0.46	0.43
OVERALL MPI
Overall MPI cut-off 0.7	0.26	0.31	0.34	0.21	0.61	0.17	0.23
Overall MPI cut-off 0.5 and 0.7	0.45	0.51	0.57	0.39	0.88	0.33	0.39
Observations	1261	407	127	727	104	87	222

The overall multidimensional wellbeing index shows that among migrant households’ 31% are well off (i.e. not deprived), compared to 34% of return migrant households and 21% of non-migrant households when a cut-off of 0.7 is used. The overall MPI is consistent with the overall differences in dimensions between the different household groups. Return migrants are the most likely to be well-off in each of the five dimensions except the inclusion dimension. It is noteworthy that return migrants are pointedly better off than not only non-migrant households, but also migrant households.

In each of the dimensions, except for education, migrant households are more likely to be better off than non-migrant households. In the education dimension, 49% of migrant households are well off as opposed to 54% of return migrant households and 50% of non-migrant households.

Comparing migrant households across migrant destination regions shows that households with migrants in the North are considerably more likely to be well-off, both for the overall multidimensional measure and when analysing each dimension separately. The largest differences across migrant households with migrants in the North compared to migrants in Africa or the Middle East are found in the housing and income dimensions. Close to 90% of households with migrants in the North are well-off in the housing dimension, compared to 21% among households with migrants in Africa and 33% among households with migrants in the Middle East. Table A1 in the appendix gives a more detailed picture of the percentage of households that are well-off for each indicator, in each dimension, and by household type. The table reveals that when it comes to differences in the income dimension between households with migrants in the North and migrants in other destination regions, the largest differences are found in income threshold ($2 per day) and the consumption asset indicators. Eighty-three percent of households with a migrant in the North achieve wellbeing in asset ownership, compared to 17% of households with a migrant in Africa and 31% of households with a migrant in the Middle East.

Another indicator that has a prominent gap between households based on migrant destination is child mortality in the health dimension. In terms of child mortality, 83% of households with a migrant in the North achieve wellbeing, compared to 47% of households with a migrant in Africa and 63% of households with a migrant in the Middle East. The only indicator upon which household with a migrant in Africa are more likely to be well-off compared to the other two groups is membership in an organization in the inclusion dimension. This is likely due to the fact that micro-credit and savings organizations are common amongst poor households in Ethiopia.

Next, a probit regression is carried out to investigate the linkages between migration and multidimensional well-being. A number of control variables on both individual and household level are included. In the first step, the link between wellbeing and migration to any destination region is investigated. In a second step, we investigate the link between migration and wellbeing across migrant destination regions. Table 4 presents the results of the first step.

Table 4. Probit regression, migration and overall MPI.

Variables	(1)		(2)
	Marginal effects	SE	Marginal effects	SE
Reference: non-migrant household Migration household	0.079***	0.022	0.027	0.026
Return migrant household	0.061*	0.035	0.061*	0.034
Household receive remittances	x	x	0.093***	0.028
Urban	0.104***	0.029	0.099***	0.029
Gender (1 = female)	−0.071***	0.023	−0.070***	0.023
Age of household head	−0.001	0.004	−0.001	0.004
Age of household head squared	−0.000	0.000	−0.000	0.000
Number of members 65 and above	−0.005	0.026	−0.003	0.026
Number of children 0–14	0.019***	0.007	0.018**	0.007
Reference: head unemployed
Head in paid work	0.053*	0.032	0.058*	0.032
Head self employed	0.084***	0.031	0.088***	0.031
Head farmer	−0.168***	0.038	−0.163***	0.037
Head retired	0.105***	0.033	0.106***	0.033
Reference: less than primary
Primary education	0.092***	0.033	0.087***	0.032
Secondary education	0.218***	0.029	0.216***	0.029
Graduate education or higher	0.419***	0.038	0.409***	0.038
Reference: Addis Ababa
Amhara	−0.020	0.034	−0.016	0.034
Oromiya	0.068**	0.033	0.068**	0.033
SNNP	0.005	0.041	0.006	0.041
Tigray	−0.104***	0.035	−0.100***	0.034
Observations	1236		1236

* significant at the 10% level

** significant at the 5% level

*** signifiant at the 1% level

The results show a positive association between migration and overall multidimensional well-being. Having a migrant increases the likelihood that a household is better off (column 1). However, this effect is only statistically significant when not controlling for remittances. When a control for remittances is introduced in the second step (column 3), there is no longer a statistically significant relationship between migration and multidimensional well-being. The link between migration and well-being therefore appears to be through monetary remittances that the migrants send.

Return migration is positively associated with household wellbeing, and this effect is robust to the inclusion of a control for remittances in column 3. This confirms the patterns in Table 3 where return migrants were systematically better off compared to migrant and non-migrant households. This may indicate that returnees bring resources with them in the return, such as financial, human, or social capital, that contribute to their household wellbeing post-return.

Table A2 in the appendix shows the results for analyses carried out for each dimension separately. The results show that the link between migration and wellbeing differ depending on which dimension that is analysed. In the education dimension, migrant households are significantly less likely to be well off than non-migrant households. This is consistent with the descriptive statistics in Table 3 wherein migrant households were slightly less likely to be well off in education than non-migrant households. Within the health dimension, migrant households are also significantly worse off, however, remittance receiving households are better off and having more migrants in the household seems to have a positive outcome. The economic dimension shows the greatest positive association with migration. Here both return migrant households and remittance receiving households are significantly better off. The housing dimension shows the weakest association with migration as none of the migration variables are significant. Finally, the inclusion dimension shows that households receiving remittances are more likely to be well off.

Several other indicators are also significant across the dimensions. In all dimensions except for health, the household head being a farmer decreases the likelihood of the household being well off. Also of significance in both the housing and inclusion dimension, female headed households are more likely to not be well off. Across all the dimensions the household head having secondary or tertiary education significantly increases the likelihood of the household being well-off. In addition, in all the dimensions except for health, urban households are more likely to be well off. The household head being self-employed had a positive association in the education, health and inclusion dimension.

In a second step, the difference in wellbeing across different migration destinations is investigated. Table 5 shows the results of the probit regression with the migration variable disaggregated into migrant destinations in Africa, the Middle East and the North.³ The results reveal a significant difference between households with a migrant in the North compared to the households with migrants in other destination regions and households without migrants. Having a migrant in the North is positively associated with household wellbeing. Households with migrants in the other destination regions are however not more likely to be better off compared to households without migrants.

Table 5. Probit regression, migration and overall MPI across different destination countries.

Variables	Marginal effect	SE
Reference: non-migrant household
Migrant in Africa	0.019	0.048
Migrant in Middle East	0.007	0.042
Migrant in the North	0.104**	0.048
Household receives remittances	0.112**	0.051
Interaction remittances and north	−0.069	0.082
Interaction Remittance and Middle East	−0.029	0.070
Interaction Remittances and Africa	−0.045	0.087
Urban	0.096***	0.030
Gender (1 = female)	−0.077***	0.024
Age of household head	−0.001	0.004
Age of household head squared	0.000	0.000
Number of members 65 and above	−0.016	0.026
Number of children 0–14	0.018**	0.007
Reference: head unemployed
Head in paid work	0.053	0.032
Head self employed	0.082***	0.031
Head farmer	−0.167***	0.038
Head retired	0.105***	0.034
Reference: less than primary
Primary education	0.092***	0.033
Secondary education	0.220	0.029
Graduate education or higher	0.398	0.038
Reference: Addis Ababa
Amhara	−0.014	0.035
Oromiya	0.079	0.033
SNNP	0.001	0.041
Tigray	−0.087	0.035
Observations	1225

** significant at the 5% level

*** significant at the 1% level

The positive association between having a migrant in the North and household wellbeing is robust and has been further tested for the inclusion of a control for remittances. It is possible that the link between remittances and wellbeing differ across migrant destination regions. In order to test this, interaction effects between remittances and the three respective destination regions are also included. None of the interaction effects are however statistically significant.

Remittance receipt is clearly associated with a high level of well-being. Here remittances not only capture income sent from any source whether from labor, social benefits or other income sources but also transnational ties between the migrant and the household. It may be logical to consider the reason for migration here but reason for migration may have nothing to do with the current activity and way of living of the migrant now so we do not include this.

Conclusion

This paper had two objectives of, first, utilizing a multidimensional wellbeing approach to assess differences between, migrant, non-migrant and return migrant households, and second, going a step further to investigate the relationship between migrant destination and well-being of the household left behind. We acknowledge that there are limitations in this approach including the arbitrariness inherent in the multidimensional methodology. While we realize the limitations of the multidimensional methodology, the strengths of the methodology maintain the opportunity to examine different dimensions of well-being, which go beyond basic poverty measures to provide a more holistic picture.

The study reveals that there is a positive association between migration and multidimensional wellbeing in Ethiopia, but only when taking the migrant destination region into account. An overall measure of migration reveals that migration is positively linked to wellbeing only when not simultaneously controlling for remittances. If remittances are taken into account, there is no longer any statistically significant association between migration and the multidimensional wellbeing measure. The link between migration and well-being is thus dependent on the household receiving remittances. We can therefore reject the first hypothesis that non-migrant households are worse off than migrant households, and confirm the second hypothesis that remittance receiving households are better off than non-migrant households.

However, when disaggregating the effect of migration according to migration destination region, the results reveal differences depending on if the household has a migrant in the North or in another destination region. Households with migrants in the North are more likely to be well-off compared to households without any migrants. The same does not hold for households with migrants in the Middle East or in other African countries. The results confirm that migrant destination is important in determining the well-being of the households left behind. Further, the results confirm hypothesis four that households with a migrant in the North are better off than households with a migrant in the South or Middle East. The results also reveal that there are differences across different dimensions of wellbeing. Migration to the North is positively associated with well-being related to education, health and inclusion, but not with measures of well-being in income or housing. The reasons for these nuances are not known.

The results also confirm that return migration is positively associated with household well-being and we can therefore accept hypothesis three. At the same time, return migration has been rapidly changing in Ethiopia since the time of this survey with increasing deportations from Saudi Arabia and the Middle East. It is important to consider if these results would still hold in a more recent survey of return migration in Ethiopia.

These findings stress the importance of not over-generalizing the positive impacts of migration to any destination. The descriptive statistics from this study illustrate that migrants to the North and the households from which they originate are better off. However, limitations in the data does not allow us to compare household well-being before and after migration, and it is therefore hard to determine ifthe higher well-being among households with migrants to the North is a result of migration or if these migrants originate from better off households to begin with.

In terms of migration to the Middle East and Africa it is important to stress that these migration patterns do not increase household well-being. As stated in the migration overview of this paper, the most prevalent migration stream from Ethiopia at this time is to the Middle East. There is no evidence that households are better-off from having a migrant in the Middle East, however at the same time, there is also no evidence that these households are worse off. Qualitative research suggests that female migration from Ethiopia to the Middle East tends to result more commonly in household survival back home than in increasing household well-being (Kuschminder, 2014). It is slightly concerning that within the descriptive statistics, households with a migrant in Africa have the lowest multidimensional well-being at 0.17, compared to all other groups, including non-migrant households at 0.21. This is not significant in the regression analysis, but still raises questions as to the drivers of migration from Ethiopia within Africa and if this migration flow is motivated for survival reasons.

Finally, we created an interaction term to test the relationship between migrant destination and remittances with the idea that households with a migrant in the North may receive more remittances. The results show these effects were insignificant and we therefore reject our final hypothesis that households with a remittance sender in the North are more likely to be better off than households with a remittance sender in the South or Middle East. These results suggest that remittances in themselves are significant, regardless of where they come from.

This paper has presented a unique comparison of the relationship between migration and household well-being with migrants in a destination in the North, Middle East, and the South. It is clear that further research needs to disaggregate the effects of migration by migrant destination. The findings in this study show that in the Ethiopian context migration to the North is very different than migration to the South and results in different levels of household well-being. Further comparative analysis is needed of South-North and North-North migration effects on the household left behind.

Funding

This work was supported by the Dutch Ministry of Foreign Affairs under the IS Academy Migration and Development Project.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Katie Kuschminder is a research fellow at the European University Institute (EUI) with research interests in irregular migration, return migration and migration and development.

Lisa Andersson is an economist at the OECD Development Centre and the EU Social Protection Systems Programme. Her research interests include migration and development, remittances, welfare and social protection.

Melissa Siegel is a professor of Migration Studies at Maastricht University/ UNU-Merit. Her research interests include migration and corruption, migration and health, and migration and development.

Acknowledgements

The authors would like to express their gratitude to the Ethiopian Development Research Institute, Asmelash Haile Tsegay, and all of the project team of enumerators for their assistance in the data collection for this study. We are also grateful to all of the respondents that took their time to participate in this study.

Notes

1. Other papers utilizing this approach include: Vanore, Siegel, Gassmann, & Waidler, 2017; Gassmann, Siegel, Vanore and Waidler, 2018; Waidler, Vanore, Gassmann, & Siegel, 2017a,b

2. Income per adult equivalence is used over income per capita to take into account the economies of scale in consumption when household members share certain goods and that children usually have lower needs than adults. The OECD equivalence scale (also called ‘Oxford scale) assigns value 1 to the first individual in the household, but less value to additional adult members and to children in the household. Alternative measures to the OECD scale (such as ‘OECD-modified scale’ and the ‘Square root scale’) assigns less value to additional members of the household besides the first adult. There is however no accepted one method for the use of equivalence scales (OECD, n.d.).

3. Table A3 in the appendix displays the analyses for the various dimensions separately.

Appendix 1

Table A1. Descriptive table, Multidimensional Wellbeing Index.

Table

Dimensions, indicators, and threshold	North	Africa		Middle East	Migrant household	Return migrant household	Household without migration experience
EDUCATION
Head of the household is literate	0.68	0.54		0.47	0.54	0.58	0.54
All children 7–15 goes to school	0.98	0.76		0.82	0.85	0.85	0.82
Education dimension (0.7 cut-off)	0.67	0.45		0.41	0.49	0.54	0.50
HOUSING
Household has access to electricity	0.95	0.75		0.44	0.56	0.65	0.59
Household has access to toilet	0.84	0.17		0.77	0.81	0.70	0.67
Household has more than one room	0.97	0.75		0.61	0.74	0.69	0.57
Floor not dirt, sand, dung	0.80	0.17		0.33	0.42	0.41	0.31
Housing dimension (0.7 cut-off)	0.88	0.21		0.33	0.46	0.47	0.35
HEALTH
Household has access to health clinic	0.96	0.97		0.99	0.97	0.98	0.96
Household able to meet its food needs (1 = yes)	0.88	0.68		0.79	0.80	0.78	0.71
Household does not have disabled or seriously ill household head	1.00	0.99		0.97	0.98	0.98	0.96
Household has not lost a child	0.84	0.47		0.63	0.64	0.74	0.73
Health dimension (0.7 cut-off)	0.99	0.80		0.90	0.90	0.90	0.89
INCOME/EMPLOYMENT/ASSETS
Household earns more than $2/day per adult equivalent	0.56	0.20	0.23		0.30	0.44	0.29
None of the children in the household are working	0.38	0.53		0.56	0.50	0.59	0.57
Household has more than one income source	0.67	0.67		0.67	0.68	0.59	0.57
Household owns at least two consumption assets	0.83	0.17		0.31	0.41	0.47	0.31
Income dimension 0.7 cut-off	0.51	0.18		0.24	0.29	0.39	0.21
INCLUSION
Household owns mobile phone	0.96	0.56		0.63	0.71	0.68	0.52
Household member of at least one organization	0.69	0.84		0.65	0.70	0.69	0.65
Inclusion dimension 0.7	0.68	0.46		0.43	0.50	0.47	0.36
OVERALL MPI
Overall mpi cut-off 0.7	0.61	0.17		0.23	0.31	0.34	0.21
Overall mpi cut-off 0.7 and 0.5	0.88	0.33		0.39	0.51	0.57	0.39
Observations	104	87		222	407	127	727

Table A2. Probit Results by Dimension.

Table

Variables	1	2	3	4	5	6	7	8	9	10
	Education dimension	se	Health dimension	se	Income dimension	se	Inclusion dimension	se	Housing dimension	se
Migration household	−0.104*	0.055	−0.082*	0.044	−0.032	0.051	−0.018	0.069	−0.043	0.045
Return migrant household (without current migrant)	−0.023	0.040	−0.012	0.028	0.137***	0.035	0.070	0.043	0.043	0.037
household receive remittances	0.030	0.034	0.054**	0.025	0.122***	0.031	0.108***	0.036	0.035	0.028
number of migrants in household	0.096**	0.038	0.072**	0.035	0.027	0.036	0.063	0.053	0.052	0.035
Urban	0.166***	0.038	−0.030	0.029	0.106***	0.035	0.087**	0.040	0.220***	0.028
Gender (1 = female)	−0.282***	0.026	−0.029	0.023	−0.029	0.027	−0.063**	0.030	−0.008	0.023
Age of household head	−0.017***	0.005	−0.010**	0.004	0.004	0.005	0.007	0.005	0.002	0.004
Age of household head squared	0.000	0.000	0.000*	0.000	−0.000	0.000	−0.000	0.000	−0.000	0.000
Number of members 65 and above	0.032	0.031	−0.010	0.025	−0.011	0.031	−0.015	0.034	−0.003	0.029
Number of children 0–14	−0.000	0.008	0.007	0.006	0.030***	0.008	0.013	0.009	0.010	0.007
Head in paid work	0.128***	0.042	−0.012	0.033	−0.012	0.038	0.031	0.044	0.006	0.032
Head self employed	0.082**	0.040	0.058*	0.033	0.058	0.037	0.104**	0.043	0.018	0.031
Head farmer	−0.097**	0.040	0.017	0.029	−0.132***	0.039	−0.078*	0.044	−0.252***	0.034
Head retired	0.144***	0.043	0.043	0.034	0.043	0.040	0.190***	0.046	0.028	0.035
Primary education	0.081**	0.033	0.028	0.022	0.067**	0.034	0.102***	0.035	0.095***	0.030
Secondary education	0.199***	0.033	0.086***	0.024	0.140***	0.032	0.235***	0.033	0.244***	0.025
Graduate education or higher	0.389***	0.052	0.199***	0.047	0.291***	0.043	0.300***	0.052	0.404***	0.042
Amhara	−0.032	0.045	−0.021	0.035	−0.049	0.039	−0.061	0.046	0.115***	0.037
Oromiya	0.049	0.045	−0.041	0.033	−0.016	0.040	0.109**	0.047	0.153***	0.034
SNNP	0.090*	0.051	−0.150***	0.037	0.035	0.046	0.098*	0.052	0.071*	0.040
Tigray	−0.021	0.045	−0.012	0.035	−0.121***	0.040	−0.175***	0.046	0.093***	0.035
Observations	1258		1249		1249		1254		1253

Table A3. Probit results with migrant destinations, by dimension.

Table

Variable	Education		Health		Income		Inclusion		Housing
Africa	0.033	0.049	−0.042	0.032	−0.043	0.049	0.056	0.052	-0.040	0.043
Middle East	−0.047	0.039	−0.002	0.026	−0.030	0.037	0.045	0.041	-0.024	0.028
North	0.096**	0.051	0.179**	0.069	0.029	0.045	0.067	0.055	0.162***	0.041
remittancehh	0.049	0.035	0.045*	0.025	0.115***	0.032	0.104***	0.037	0.026	0.028
Urban	0.157***	0.038	−0.028	0.028	0.113***	0.035	0.108***	0.040	0.218***	0.028
Gender (1 = female)	−0.277***	0.026	−0.034	0.022	−0.038	0.027	-0.068**	0.030	-0.010	0.023
Age of household head	−0.017***	0.005	−0.009**	0.004	0.005	0.005	0.007	0.005	0.001	0.004
Age of household head squared	0.000	0.000	0.000*	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Number of members 65 and above	0.029	0.030	−0.019	0.025	−0.009	0.032	-0.013	0.035	-0.011	0.028
Number of children 0–14	0.000	0.009	0.006	0.006	0.030	0.008	0.013	0.009	0.011	0.007
Head in paid work	0.135***	0.042	−0.013	0.032	−0.022	0.039	0.027	0.044	0.012	0.032
Head self employed	0.089**	0.040	0.057*	0.032	0.059	0.037	0.105**	0.043	0.026	0.030
Head farmer	−0.095**	0.041	0.022	0.029	−0.127***	0.040	-0.073*	0.044	-0.227***	0.034
Head retired	0.137***	0.044	0.035	0.034	0.040	0.041	0.189***	0.047	0.021	0.034
Primary education	0.079**	0.033	0.031	0.022	0.067*	0.034	0.100***	0.035	0.092	0.030
Secondary education	0.197***	0.033	0.081***	0.024	0.144***	0.032	0.231***	0.034	0.238***	0.025
Graduate education or higher	0.372***	0.053	0.177***	0.048	0.292***	0.044	0.298***	0.053	0.370***	0.042
Amhara	−0.039	0.046	−0.020	0.036	−0.055	0.041	-0.044	0.047	0.105***	0.038
Oromiya	0.045	0.045	−0.034	0.033	−0.006	0.040	0.119**	0.047	0.158***	0.033
SNNP	0.086*	0.051	−0.145***	0.037	0.026	0.047	0.109**	0.053	0.064	0.040
Tigray	−0.017	0.045	−0.006	0.035	−0.112***	0.040	-0.166***	0.046	0.100***	0.034
Observations	1244		1235		1235		1240		1239