Multidimensional poverty and the variables that contribute to it in the Gurage Zone of Ethiopia: Using rural-urban migration as a decomposition parameter

Abstract

BACKGROUND:

Poverty alleviation is a critical issue for national and international development goals. Data on different poverty-reduction measures are eagerly sought by policymakers. Analytical data on the role of migration in poverty alleviation is one of them.

OBJECTIVES:

This study is designed to investigate the prevalence of multidimensional poverty and its determinants in connection to rural-urban migration.

METHODS:

Primary data was collected from 384 randomly selected rural households and analyzed using econometric models.

RESULTS:

Non-migrant and migrant-sending households had adjusted headcount ratios of 19.8% and 10.5%, respectively. Poor living conditions were identified in 43.5% of non-migrant households and 25.6% of migrant-sending households, respectively. Non-migrant households and migrant-sending households contributed 70.5% and 29.5%, respectively, to the entire sample’s adjusted headcount ratio. According to the findings, household size, the number of migrants per household, the education level of the household head and livestock ownership all has a significant effect on households’ multidimensional poverty.

CONCLUSIONS:

The results demonstrate that rural-urban migration helps poverty reduction in the region. Therefore, proper consideration should be given to maximizing the benefit of migration on the long-term reduction of multidimensional poverty through productive remittance investment and assisting households to improve their productive capacity.

Keywords

Determinants multidimensional poverty rural-urban migration Ethiopia

Mesfin Agza is a PhD candidate at Addis Ababa University in Ethiopia. He is also a lecturer in the department of Agricultural Economics at Wolkite University and a corresponding author of this work. ORCID: 0000-0002-9681-8717.

Bamlaku Alamirew is an Associate Professor at Yom postgraduate college and a Certified International Project Manager (CIPM) in Ethiopia.

Admasu Shibru is an Associate Professor at the Center for Food Security Studies, Addis Ababa University, Ethiopia.

1 Introduction

Multidimensional poverty and rural-urban migration are conceptually linked, particularly in developing countries. Rural-urban migration as a livelihood strategy has a dual impact on the lives of migrant sending households, depending on their management abilities. First, rural-urban migration is a source of remittance money that helps to alleviate poverty by providing basic commodities, education, and health care [1, 2]. Second, migration is blamed for the causes of family brain and skill drain, which leads to a loss of economically engaged and skilled persons, followed by multidimensional poverty [3].

Long-term rural-urban migration as a livelihood strategy has been well-cultivated in Ethiopia, notably in the Gurage zone, and it is on the rise and is expected to continue in the future [4 –6]. Despite the fact that there has been a significant amount of empirical research in Ethiopia on migration and household livelihood as separate concerns, only a few studies [7, 8] have looked at how they interact with one another. These few studies are also focused on evaluating welfare outcomes and do not give a full picture of multidimensional poverty. Furthermore, with the exception of a few research studies [9, 10], using a comprehensive technique to assess poverty is uncommon in Ethiopia. To the best of the researchers’ knowledge, earlier investigations in the Gurage zone on the research topic are quite limited in scope and require a quantitative analysis to make a conclusion through hypothesis testing. This study aims to fill the gaps mentioned above by providing analytical data on multidimensional poverty in households, the link between migration and poverty, and the causes of poverty. The study’s methodology might serve as a foundation for future similar studies, and policymakers could use the analytical data supplied when designing a poverty-fighting strategy.

1.1 Theoretical background

Poverty has been studied and explored from several angles. Migration is one of a comparative parameter, and various studies have provided analytical data on it and other factors that contribute to poverty. According to Zhang [11], household poverty is a driving force behind rural–urban migration, and anti-poverty work had also played a significant role in developing rural–urban migration adjustment system in China. Tanle, Ogunleye-Adetona, and Arthor [12] examined migration status to break down poverty and found that migrant-sending households had better access to health care, more educational possibilities, and larger agricultural earnings than non-migrant-sending families. Similarly, Kuschminder, Andersson, and Seigel [8] used related characteristics to compare well-being and discovered that migrant households outperform non-migrant ones. Sanusi et al. [13] classified poverty into rural and urban subgroups and investigated the factors that contribute to it in India. They reported a high level of household poverty, with rural families being more likely than urban families to be impoverished. Sanitation, cooking fuel, child health, housing conditions, and the family head’s education were all elements that contributed to the household’s multidimensional poverty. Similarly, Megbowon [14] investigated multidimensional poverty and its determinants in South Africa’s Eastern Cape Province, then compared the findings between urban and rural households. He reported a higher adjusted poverty headcount ratio in rural areas, which was significantly influenced by a household head’s education, access to electricity, and asset stock in both geographic areas. Dehury and Mohanty [15] also reported that the provision of sanitation, drinking water, and cooking fuel enhanced livelihood security, which in turn reduced migration. In addition, Wang and Wang [16] identified unsafe housing, family health, adult illiteracy, children enrollment rate, and fuel type as factors contributing to poverty. Sulaimon [17] looked at multidimensional poverty and broke it down by geopolitical area in Nigeria. He identified substantial variations in multidimensional poverty between the southern and northern regions, as well as the majority of northern sub-regions. He asserted that labor force participation and fertility rates have a major influence on multidimensional poverty, with the former having a negative link.

Furthermore, Amao, Ayantoye, & Fanifosi [18] investigated household multidimensional poverty and decomposed the results by geopolitical zones. In their study, they discovered an adjusted (censored) headcount ratio of 41%, while living circumstances, education, health, and assets were all recognized as key drivers to poverty. Nguyen [19] also conducted research to identify the determinants of a household’s multidimensional poverty in Vietnam’s Khmer ethnic minority. He reported that the occupation of the household head, educational level, dependency ratio, involvement in health insurance, and communication services are all important factors in household multidimensional poverty, and that all of the identified factors, with the exception of dependency ratio, were inversely related to household multidimensional poverty. Gebrekidan, Bizuneh, and Cameron [20] investigated multidimensional poverty and the factors that contribute to it in northern Ethiopia. Their findings revealed that 60% of the households were multidimensionally poor, and that socioeconomic factors like extension contact, the education of the head of the household, family size, plot size, annual household income, and access to hired labor all had a negative impact on multidimensional poverty. Maity & Buysse [21] also calculated multidimensional poverty and identified its determinants. They reported a greater degree of poverty, which is influenced by health, literacy, work possibilities, and monthly household consumption spending. Fonta et al., [22] used children aged 5 to 18 to explore multidimensional poverty, its causes, and decomposition. They observed that children who grew up in households with an intellectually minded mother, were teenage, and resided in cities had a lower risk of experiencing multidimensional poverty. Children from polygamous households, households with a head suffering from a long-term sickness, debt-ridden families, and households with more than five children were the lowest. According to Dereje, Abrham, and Alemu [23], a slight increase in household resilience capacity resulted in a consistent decrease in multidimensional poverty in Ethiopia. Meekaew and Ayuwat [24] identify factors affecting the livelihood security of migrant households in Thailand. They reported that landholding size, household assets, and capital such as human, social, physical, and natural capital are important factors that have a direct relationship with households’ livelihood security.

In general, prevailing assumptions about the relationship between migration and household poverty are frequently disputed, with results differing depending on the research and area. This section summarizes the findings of previous studies, particularly in Ethiopia and the research region, with an emphasis on major factors addressed, methodologies used, and conclusions relevant to the implications of rural-urban migration on poverty and the determinants of multidimensional poverty in households.

2 Methodology

2.1 Description of the study area

Gurage zone is one of the zonal administrations in the Southern Nations, Nationalities, and Peoples Regional State, located in the country’s southwestern corner and on the region’s northern edge between 7 ⁰ 40’ to 8 ⁰ 30^′ North and 37 ⁰ 30^′ to 38 ⁰ 40^′ East (Fig. 1). The zone comprises 5932 km² and the bulk of the land area has been significantly deteriorated, notably non-homestead agricultural lands [25, 26]. Agriculture employs 95 percent of the population, even though its contribution to household income varies by agro-ecological zone and is declining, particularly in the zone’s midlands, which account for 65.3 percent of total land area. Many households are seeking for alternative sustainable livelihood options, such as rural-urban migration, as a result of the conventional agriculture sector in the study region not offering adequate employment and income. Internal migration has always existed in permanent or seasonal patterns, benefitting both people and communities [6, 27]. With these push and pull dynamics in place, as well as recommendations from local indigenous people, rural-urban migration may become more widespread in the future, perhaps forcing many people to abandon their family, homes, land, and other assets.

Fig .1

Map of the research area. Source: Adopted form [26].

2.2 Sample and sampling techniques

The study used a multistage sampling technique to draw sample for the study. The Gurage zone is purposely chosen for the first stage of sampling since it is renowned for significant migration of individuals from rural portions of the zone to different cities and towns in Ethiopia. Following that, the districts of Gurage zone were classified as lowland, midland, and highland based on agro-ecological characteristics. Then, from the classified strata, three districts are chosen using a simple random selection technique (one district from each stratum) under the premise of agro-ecological homogeneity. Following that, six rural Kebele administrations were picked at random from the three districts. Finally, the study employed a random sampling technique [28] to choose sample households from the total rural households in the sample districts based on the sample size proportionality concept. Because we don’t know how many households consider rural-urban migration a feasible livelihood alternative, the sample size was determined using the Cochran [29] method. It used a proportionate value of 0.5 for maximum variability (p), i.e., q or 1-p=0.5; a 95% confidence level; a 5% error (e); and a 1.96 Z-value calculated from statistical tables. Consequently, the sample size is projected as: $n = \frac{(z^{2} * p * q)}{e^{2}} = \frac{^{{1.96}^{2}} * 05 * 05}{{0.05}^{2}} = 384$ (1)

2.3 Indicators and measurements

The study surveyed 384 randomly selected sample households, employing three core HDI components among the 10 publicly announced HDI dimensions. These are the dimensions of education, health, and living standards [30]. According to the approach of multidimensional poverty analysis, relevant indicators with their normative thresholds or cutoffs were established for the identified dimensions. The poor households were identified using a two-cutoff procedure, i.e. the first set of cutoffs determines whether a person is deprived across all dimensions, but the second set (a single poverty line) identifies a person as poor, [31, 30]. To distinguish the multidimensional impoverished households, a normative dimension cutoff (33%) was created by dividing 1 by the desired number of dimensions [30]. Second, to determine the poor’s deprivation score, the indicators’ weights or indicator cutoffs (poverty line) were calculated by dividing 1/3 by the number of indicators in each dimension. Each deprivation score was calculated using a weighted sum of the number of deprivations, with the goal of keeping the overall deprivation score for each household between 0 and 1. The score increases as the household’s deprivations increase, reaching a maximum of 1 when the household is deficient in all component indicators and 0 when the household is not deficient in any indicator [31, 30]. The MPI (multidimensional poverty index) is derived from the ten indicators, with a weighted vector of 1.67 for health and education indicators and 0.56 for the living standard indicators (Table 1). The indicator weight was calculated by dividing the dimension cutoff [33 percent) by the number of indicators in each dimension. Table 1 also shows the SDG (sustainable development goals) categories rwith the third goal being good health and wellbeing rthe fourth goal being education rSDG6 being clean water and sanitation rSDG7 being affordable and clean energy rSDG9 being industry rinnovation rand infrastructure rand SDG11 being sustainable cities and communities.

Table 1
Indicators and measurement

Dimensions Indicators The household is not deprived if. . . Related to. . . Weight of indicators

Education Literacy No illiterate person (>15 years) SDG4 0.167

Schooling No child (7-16 age) out of school SDG4 0.167

Health Good health No person with chronic illness SDG3 0.167

Medical Able to meet medical needs SDG3 0.167

Living Room More than one room SDG11 0.056

Standard Electricity Access to electricity (at least solar) SDG7 0.056

Water Access to improved water SDG6 0.056

Sanitation Access to private toilet SDG6 0.056

Fuel Energy source for cooking other than charcoal SDG7 0.056

Communication/asset Access to more than one communication/media assets SDG9 0.056

Dimensions	Indicators	The household is not deprived if. . .	Related to. . .	Weight of indicators
Education	Literacy	No illiterate person (>15 years)	SDG4	0.167
	Schooling	No child (7-16 age) out of school	SDG4	0.167
Health	Good health	No person with chronic illness	SDG3	0.167
	Medical	Able to meet medical needs	SDG3	0.167
Living	Room	More than one room	SDG11	0.056
Standard	Electricity	Access to electricity (at least solar)	SDG7	0.056
	Water	Access to improved water	SDG6	0.056
	Sanitation	Access to private toilet	SDG6	0.056
	Fuel	Energy source for cooking other than charcoal	SDG7	0.056
	Communication/asset	Access to more than one communication/media assets	SDG9	0.056

Source: Own survey result, 2021.

2.4 Methods of data analysis

The prevalence of rural household poverty, determinants of multidimensional poverty, and the impact of migration on multidimensional poverty in the study area were investigated using a multidimensional poverty index, probit regression model, and propensity score matching. The brief description of each model is provided below.

2.4.1 Specification of multidimensional poverty index

The multidimensional poverty index takes a comprehensive approach to assessing poverty by considering aspects such as education, health, and social inclusion in addition to monetary factors 32–34. The index uncovers the deprivations that the poor households face, as well as the links among the elements of deprivations. In the estimation process, the generalized version of Alkire and Foster [31] index was used with the help of DASP software. The index is defined as follows, according to Alkire et al. [30]: $P (X, Z) = \frac{\sum_{i = 1}^{n} w_{i} p (X_{i}, Z)}{\sum_{i = 1}^{n} w_{i}}$ (2)

Where p (X_i, Z) is the household poverty level (with the vector of indicators X_i = (X_i,1 … X_i,j) and the vector of poverty lines Z = (Z . . . . . . Z); and the contribution of indicators to total poverty is p = (X, Z). Furthermore, formula 1 and 2 are used to calculate the relative contribution of population subgroups and indicators to multidimensional poverty indices. $C_{t}^{o} = S^{p} \frac{M_{O} (X^{t})}{M_{o} (X)}$ (3)

Where $C_{t}^{0}$ is the subgroup’s poverty contribution, S^P is the subgroup’s sample population share, M_o (X^t) is the subgroup’s MPI, and M_O (X) is the population’s MPI. The following formula was used to calculate the contribution of indicators to overall poverty M_O: $φ_{i}^{0} (k) = \underset{i}{w} \frac{h_{i} (k)}{M_{O}}$ (4)

Where $φ_{i}^{o} (k)$ is the contribution of an indicator at a dimension cutoff point k, W_i is the indicator’s weight, $\underset{i}{h} (k)$ is the censored headcount ratio at an indicator I a cutoff point k, and M_O is the population’s adjusted headcount poverty ratio.

2.4.2 Specification of probit model

By categorizing households as poor or non-poor based on their censored deprivation score Ci; the binary choice model was used to determine the key factors of multidimensional poverty in the research area. Households with a censored deprivation score (Ci) greater than the poverty criterion (33%) were classified as multidimensional poor. As a result, the outcome variable (y) is treated as a dummy variable, with values of 1 and 0 representing multidimensional poor and non-poor households, respectively, and a dimension cutoff point k. It is written as follows: $Y_{i} {\begin{matrix} 1 & if Ci \geq k \\ 0 & otherwise \end{matrix}$ (5)

After that, the maximum likelihood estimator was used to fit the probit regression model to estimate the chance of a household being multidimensional poor (y = 1) [35, 36], and the findings were interpreted in terms of marginal effect [37]. As a result, the model is defined as follows: $y_{i} = {\begin{matrix} 0 & if y^{*} ⩽ 0 \\ 1 & if y^{*} \geq 0 \end{matrix}$ (6)

Where y_i is a dependent variable with a binary form that takes 1 for multidimensional poor households and 0 for non-poor a household as a base category, and y^* is a latent variable with y^* = Xβ + ɛ, € ∼ N (0, δ²). X and β are vectors of explanatory factors and unknown parameters, respectively, computed using maximum likelihood estimation methods. On the other hand, € is a random disturbance phrase.

Non-indicator measurement variables were used as explanatory variables in the model. This comprises migration, which is a dummy variable with a value of 1 if at least one member of the household is a migrant and a value of 0 otherwise; household size; and household head’s education level, which is defined as the number of years of schooling; sex of household heads, represented by a dummy variable with a value of 1 if the household head is a male and zero otherwise; livestock holding, defined as the number of livestock in TLU; and soil infertility, represented by a dummy variable with a value of 1 if the household experienced soil infertility and zero otherwise.

3 Results and discussion

This section presents the results of the household survey data analysis focusing on the impact of migration on multidimensional poverty. In the preliminary part, the various measures of multidimensional poverty are presented by disaggregating for the two household groups classified based on their migration status. The subsequent sections explain the determinants of multidimensional poverty and the impact of migration on MPI in the study area that are estimated using econometric models.

3.1 Raw headcount deprivations of the rural households in the study area

Contingency Table 2 demonstrates the measure of raw headcount deprivation in each indicator for the household groups. The table gives a picture of the proportion of households that are either well-off or deprived for each indicator. The deprivation of the whole sample households in child schooling and adult literacy was 11.5 and 49%, respectively. Migrant-sending households experienced more deprivations in the education dimension. Of the deprivations of all households in education, 86.36% of the deprived population in children’s schooling and 58% in adult literacy were shared by migrant-sending households. The remaining 13.64% of the deprived population in children’s schooling and 42% in adult literacy were non-migrant households. In the study area, the deprivation of rural households in the health dimension was relatively lower than the other two dimensions. Excluding the number of people in the household, 5.73 percent and 16.67 percent of the sample households were deprived of good health and medical needs, respectively. Regarding the deprivations of the two subgroups in good health and medical needs, the deprivations for migrant-sending households are 18.2 percent and 23.4 percent, respectively, while the proportions for non-migrant households are 81.8 percent and 76.6 percent. This study found that the average deprivation of sample households in all indicators of living standards was very low, with the exception of indicators such as access to electricity (46.1 percent) and energy sources for cooking other than solid fuel (76.56 percent). Households with migrant members were significantly more likely to be better off in all indicators of living standards except access to an energy source for cooking, regardless of the number of people in the household. In which they experienced 55.8 percent of the deprivation brought on by a lack of cooking fuel. The number of rooms, access to more than one communication or media asset, and access to a household level toilet were found to be the most significant differences between migrant-sending households and non-migrant households. Non-migrant households exhibited a far larger proportion of deprivation than migrant-sending households in these parameters; their deprivation share in the aforementioned categories was significantly greater than 50%. Keep in mind that the raw headcount deprivation mentioned in Table 2 was analyzed without taking the number of individuals in the household into account; therefore, non-poor households were included in the analysis.

Table 2
The extent of multidimensional deprivations experienced by household groups

Dimensions Indicators Deprivation status Household subgroups

Migrant-sending Non migrant Total

(214) (170) (384)

Education Child schooling Non-deprived=0 176 164 340

Deprived=1 38 6 44

Adult literacy Non-deprived=0 105 91 196

Deprived=1 109 79 188

Health Good health Non-deprived=0 210 152 362

Deprived=1 4 18 22

Meeting medical needs Non-deprived=0 199 121 320

Deprived=1 15 49 64

Access to above one room Non-deprived=0 208 129 337

Deprived = 1 6 41 47

Living standards Access to electricity Non-deprived=0 125 82 207

Deprived=1 89 88 177

Access to drinking water Non-deprived=0 170 116 286

Deprived=1 44 54 98

Access to toilet Non-deprived=0 203 139 342

Deprived=1 11 31 42

Access to cooking fuel Non-deprived=0 50 40 90

Deprived = 1 164 130 294

Communication asset Non-deprived=0 193 105 298

Deprived = 1 21 65 86

Dimensions	Indicators	Deprivation status	Household subgroups
Education	Child schooling	Non-deprived=0	176	164	340
		Deprived=1	38	6	44
	Adult literacy	Non-deprived=0	105	91	196
		Deprived=1	109	79	188
Health	Good health	Non-deprived=0	210	152	362
		Deprived=1	4	18	22
	Meeting medical needs	Non-deprived=0	199	121	320
		Deprived=1	15	49	64
	Access to above one room	Non-deprived=0	208	129	337
		Deprived = 1	6	41	47
Living standards	Access to electricity	Non-deprived=0	125	82	207
		Deprived=1	89	88	177
	Access to drinking water	Non-deprived=0	170	116	286
		Deprived=1	44	54	98
	Access to toilet	Non-deprived=0	203	139	342
		Deprived=1	11	31	42
	Access to cooking fuel	Non-deprived=0	50	40	90
		Deprived = 1	164	130	294
	Communication asset	Non-deprived=0	193	105	298
		Deprived = 1	21	65	86

Source: Own survey result, 2021.

3.2 The scope of multidimensional poverty analysis

Table 3 shows estimates of household multidimensional poverty coverage at a poverty threshold of k = 33%. The share of the sample population in the analysis of MPI is 44.1% and 55.9 % in migrant-sending and non-migrant households, respectively. The MPI is more detailed than headcount deprivation since it includes information on the population’s proportion of weighted deprivation. We use another category, poor and non-poor, because these two groups were present in both migrant-sending and non-migrant households, and we want to compare two poor categories in the sample. The MPI of the two subgroups was estimated by censoring the deprivations of the non-poor and calculating the proportion of people who have been recognized as multidimensionally poor in the population. The value of MPI was broken down into the incidence and intensity of poverty since it is a linear product of the two indices.

Table 3
The respondents’ multidimensional poverty indices (observation 384)

Population Pop. MPI Poverty Intensity of Contributions to. . .

subgroups Share value headcount deprivation MPI Headcount Intensity

NMHHHs 0.559 0.198 0.455 0.435 0.705 0.693 0.568

MSHHs 0.441 0.105 0.256 0.410 0.295 0.307 0.432

Whole sample 1.000 0.157 0.367 0.428 1.000 1.000 1.000

Difference –0.118 –0.093 –0.199 –0.025 –0.41 –0.386 –0.136

Population	Pop.	MPI	Poverty	Intensity of	Contributions to. . .
NMHHHs	0.559	0.198	0.455	0.435	0.705	0.693	0.568
MSHHs	0.441	0.105	0.256	0.410	0.295	0.307	0.432
Whole sample	1.000	0.157	0.367	0.428	1.000	1.000	1.000
Difference	–0.118	–0.093	–0.199	–0.025	–0.41	–0.386	–0.136

Note: The difference is calculated between MSHHs (migrant-sending households) and NMHHHs (non-migrant households). Source: Own survey result, 2021.

The multidimensional poverty index of subgroups is obtained by adding the censored deprivation scores weighted by the population share of each household group or by multiplying poverty headcount ratio by the intensity of poverty (0.367*0.428 = 0.157). Following poverty decomposability property, the entire multidimensional poverty index of the study area could be also obtained by adding the poverty level of subgroups after multiplying by each sample population shares’, which is estimated as 0.559 * 0.198 + 0.441 * 0.105 = 0.157. As indicated in Table 4, the study satisfied dimensional monotonicity property [30, 34] as a poor person became deprived in extra indicators, then the intensity of poverty and MPI values were rising together.

Table 4

The proportion of poor individuals and those facing deprivation in each indicator

Indicators	Weight	Censored headcount ratio of indicators			Percent Contribution of indicators to MPI(WiHR/M0)=0.157*
		Non-migrant HHs	Migrant HHs	All HHs	Non-mig HHs	Migrant HHs	All HHs (0.157)
Schooling	0.167	0.043	0.101	0.0683	4.63	10.74	7.2
Literacy	0.167	0.2996	0.2104	0.2603	31.87	22.38	27.62
Good health	0.167	0.101	0.0193	0.0649	10.74	2.05	6.90
Medical	0.167	0.2385	0.0701	0.1638	25.37	7.46	17.43
Room	0.056	0.1574	0.0363	0.1031	5.61	1.30	3.70
Electricity	0.056	0.3378	0.1898	0.2739	12.91	6.77	9.77
Water	0.056	0.2328	0.1125	0.1791	8.30	3.98	6.39
Sanitation	0.056	0.1116	0.0302	0.0754	3.98	1.10	2.69
Fuel	0.056	0.4169	0.2455	0.3403	14.87	9.05	12.14
Media assets	0.056	0.2538	0.0677	0.1713	9.06	2.40	6.11
Total	1	2.1924	1.0828	1.7004			100

Source: Own survey result, 2021.

As shown in Table 4, there is a greater disparity between the two subgroups in terms of population share, multidimensional poverty index, and poverty headcount. Non-migrant households’ multidimensional poverty, for example, virtually quadrupled since the linked value of headcount and intensity is higher for them. The difference in the intensity of deprivation (the average deprivation share or the breadth of poverty) between the two subgroups, on the other hand, was not significant. In which the difference in weighted indicators between non-migrant and migrant-sending households was 2.5 percent (43.5 percent–41 percent). The contribution of non-migrant and migrant-sending households to the overall multidimensional poverty was 70.5% and 29.5 %; respectively. This suggests that there is an unequal distribution of multidimensional poverty [8, 30] between the migrant-sending and non-migrant household groups. The results revealed that the poverty contributions of non-migrant household group in the overall MPI, poverty headcount and intensity of deprivation are higher by 41%, 38.6% and 13.6%, respectively than that of the migrant-sending household group.

3.3 The relative contribution of different indicators to the overall multidimensional poverty

Table 4 presents the multidimensional poverty index of the ten indicators that are grouped into three dimensions. The weights were computed such that each dimension receives an equal weight of 1/3 and the weight is equally divided by the number of indicators in each dimension. As a result, each education and health indicator received larger weights (0.167) than the standard of living indicator (0.056). The table also displays the relative contribution of indicators to the aggregate deprivations of poverty (censored headcount). The contribution of an indicator to poverty has a key message to the proportion of the population who are deprived in that indicator. Using the value of headcount ratio and the weight of each indicator, the contribution of each indicator is computed. Even though, the headcount ratio of an indicator can be computed from uncensored (raw) deprivation matrix (aggregate deprivation of poor and non- poor) and censored deprivation matrix [30], only the censored headcount ratio is used in the estimation process.

Table 4 shows that the proportion of poor people who are destitute in each indicator referring the multidimensional poverty index. We clarify that these are ‘multidimensional’ poor people as those people who are deprived in 33% and above. However, the rest were considered as non-poor since they have lower proportion of deprivations. Looking at the censored headcount rations, we can see that the poor people in the study area exhibit the higher deprivation levels with respect to access for energy source for cooking and literacy level, followed by the ability to meet medical needs, and access to electricity. When an indicator’s percentage contribution to aggregate poverty surpassed its weight, a larger censored headcount ratio was found. Adult literacy, medical needs, access to electricity, access to drinkable water, and access to energy sources other than solid fuel have higher percentage contributions than their weight, and thus are associated with a higher censored headcount ratio (see the Table 4). As a result, impoverished persons are more likely to be deficient in these indicators, which are policy important factors for addressing the composition of multidimensional poverty in the research region. It was noticed that fuel and electricity have lower contribution to the overall multidimensional poverty, even though their censored headcount ratio is greater than any other indicators. This is because the weights assigned to these indicators are lower than those assigned to literacy and medical needs. In migrant-sending households, adult literacy, child schooling, access to energy source for cooking other than solid fuel, and the ability to meet medical needs were the top four indicators as they collectively contribute to 49.63% of the deprivations. However, for non-migrant household population, adult illiteracy, ability to meet medical needs, access to energy source for cooking other than solid fuel and access to electricity were the top four deprivations with 84.5% contribution to the overall multidimensional poverty. Comparing the two subgroups, except in child schooling, people in migrant-sending households were better off in all indicators which is consistent with the study result of Kuschminder, Andersson, & Seigel [8]. Generally, in the study area the contribution of education, health and living standard dimensions to the overall multidimensional poverty index is 34.8%, 24.3% and 40.9%, respectively. Oshio and Kan [38] discovered a poor education as a significant poverty component for smoking, which is similar to the findings of this study.

3.4 The determinants of multidimensional poverty

The poverty level of rural-households is associated with various demographic and socioeconomic characteristics in the study area. In this section, sample households are categorized as poor (1) or non-poor (0) based on the censored deprivation score (Ci) value in order to discover the determinants of multidimensional poverty. Households with a censored deprivation score (Ci) greater than or equal to the poverty line (33%) are classified as multidimensional poor, whereas those with a (Ci) less than the poverty line (33%) are classified as non-poor. To assess the determinants of multidimensional poverty, a probit regression model was employed, using households’ poverty status (poor or non-poor) as a dependent variable. Eleven explanatory (6 continuous and 5 discrete) variables were identified and tested for their significance of association with household poverty status. Of which, six explanatory variables (such as household size, number of migrants, age and education of household head, livestock holding, and soil infertility) were found to be significantly associated with, and chosen for the model specification.

Table 5 presents the estimated results, i.e., the coefficients, marginal effect, and associated p-values of the coefficients of the regression model. The log likelihood –210.70697 with a p-value of 0.000 indicated that the model as a whole is statistically significant and fits better than a model with no predictors. As the probit regression model output revealed, the coefficients of the number of migrants and livestock holding in TLU was negative and has an inverse relationship with the household multidimensional poverty. The explanatory variables such as gender and education of the household head, household size and soil infertility problem have positive relationship with household multidimensional poverty. Except sex of the household head and soil infertility problem all the other explanatory variables were statistically significant and determine the poverty of rural households.

Table 5
The outcome of the probit model for the factors of multidimensional poverty

Explanatory variable Coefficients Std. Err Z-value P-value Marginal effect

Household size 0.1738669 0.08683 3.43 0.001*** 0.0541605

Number of migrants –0.0900571 0.0841488 –1.84 0.065* –0.0280533

Education of HH head 0.0534579 0.0301489 2.98 0.003* 0.0166524

Gender of HH head 0.2911225 0.357467 1.40 0.160 0.0906862

Livestock holding –0.1064683 0.0748192 –2.63 0.009 –0.0331654

Soil infertility .0252202 0.2538265 0.17 0.867 .0078562

Constant –1.414831 0.651115 –3.65 0.000***

Explanatory variable	Coefficients	Std. Err	Z-value	P-value	Marginal effect
Household size	0.1738669	0.08683	3.43	0.001***	0.0541605
Number of migrants	–0.0900571	0.0841488	–1.84	0.065*	–0.0280533
Education of HH head	0.0534579	0.0301489	2.98	0.003***	0.0166524
Gender of HH head	0.2911225	0.357467	1.40	0.160	0.0906862
Livestock holding	–0.1064683	0.0748192	–2.63	0.009**	–0.0331654
Soil infertility	.0252202	0.2538265	0.17	0.867	.0078562
Constant	–1.414831	0.651115	–3.65	0.000***

Number of Obs = 384; LR chi2(6)=58.69; Prob > chi2 = 0.0000; Pseudo R² = 0.1222. Where ***, ** and * are significant levels at 1%, 5% and 10%, respectively. Source: Own survey result, 2021.

As the econometric model revealed, having a migrant household member decrease the possibility of rural household multidimensional poverty and the effect is statistically significant at 10%. The marginal effect of the number of migrants in the household indicated that additional number of migrants decrease the likelihoods of being multidimensional poor by 2.8%, holding other variables constant. The result is in line with various optimistic studies, which argue remittance-receiving households are better off, and having more migrants in the household seems to reduce multidimensional poverty through smoothing household income and increasing access to capital [8 , 40]. Besides, the neoclassical migration theory highlights migration as an investment where the benefits gained from migration has to exceed the costs associated for migration to take place. The gains obtained from migration are a flow of remittances, skills, knowledge, experience, and other household amenities that migrants acquired and are expected to be used in migrant-sending households [6 , 42]. This finding is in contradiction with the structural migration theory that argued the multidimensional poverty of rural-households increased due to the effect of migration. Migration pessimistic studies signified that the remittance obtained from migrants creates remittance dependent society and cannot cover all the costs incurred by migrant-sending households as spent on conspicuous consumption and unproductive investments [43, 44].

The size of the household was also another liable factor for determining the level of multidimensional poverty in the study area. In this research, strong positive relationship was found between household size and multidimensional poverty. While the household size is increases, all the measures of multidimensional poverty levels also increase at 1% significance level. The result of the study shows that for each additional household member the likelihood of rural household multidimensional poverty increase by 5.42%, ceteris paribus. This is because household with larger family size are challenged to meet all the necessary requirements for life. It is obvious that larger household size has acutely rooted in the poverty circle since larger households are required with higher levels of income and other household amenities to live [30, 45].

The education of household head is significantly related to the multidimensional poverty of households at a significance level of 1% and shows unexpected sign. The study signified that, for a given household head a unit increase in the year of schooling would enlarge the probability multidimensional poverty by 1.67%. An allusion of this is that households with highly educated head have a higher possibility of being poorer than their counterparts. The possible reason for this may be household heads with higher year of education are in dilemma to engage in petty works and thus gain lower income. Consequently, they are more likely to be poor. In the study area, there are many jobs that are not suitable for the educated person, maybe for their risk or for the societal norms and values. This result is inconsistence with the study result of [30 , 46] which reported that an increase in the year of education is a signal for the household to engage in different livelihood activities and it creates higher access to information that will help households to improve their way of life.

The correlation between livestock holding and multidimensional poverty of rural household is indirect and statistically significant at 5%. The marginal effect of TLU showed that a unit increase in TLU decreases the probability of a household to be multidimensional poor by 3.3%. This is because, households with a greater TLU have provided with a wide spectrum of benefits, such as cash income, food, manure, draft power and transportation services, savings and insurance, and social status and social capital [47, 48] which are basic for reducing household poverty. This finding is verified the outcome of [49 –52].

4 Conclusions and recommendations

Poverty is the greatest severe danger to the economy and people of emerging nations like Ethiopia. A proportion of rural households in the study area perceive rural-urban migration as a means out of poverty. According to the findings of this study, rural-urban migration has a considerable influence on decreasing multidimensional poverty among rural households. The contribution of rural households to overall multidimensional poverty varies greatly depending on migration status. Except for child education, migrant-sending households outperformed with all other indicators of multidimensional poverty. This study confirmed that the number of migrants, household size, household head’s education level, and livestock ownership are the major determinants of multidimensional poverty in the study area. Development planners must mainstream rural-urban migration and focus on intensity rather than on the headcount ratio so as to reduce multidimensional poverty more effectively. It is critical for poverty-reduction efforts to encourage potential migrants to provide tempting livelihood possibilities and to invest in overlapping deprivations. To do this, institutions should aggressively promote and support migrants by giving land for investment in rural regions in order to enhance remittance productivity and expand its effect on the community level by developing non-farming businesses. If this is the case, non-migrant households are also more likely to benefit since it expands a range of income options and market prospects in the sending locations. This is also because, in many circumstances, people and products movements have a favorable influence on the development of prospective local markets on the receiving end. This study also invites stakeholders to actively participate in societal training and awareness-raising about the factors highlighted in the study as contributors to multidimensional poverty in the study area.

Footnotes

Acknowledgments

Local government officials, data enumerators, and sample households in the research area, as well as Addis Ababa University, all participated in the study by giving critical information. The authors would like to thank all of the collaborators for their outstanding efforts.

Author contributions

All of the authors contributed to the development of the paper and agreed to submit it for publication.

Statement on data availability

The authors have developed datasets to support the study’s conclusions, which are available via the following link: .

References

Hidayati

Migration and rural development: The impact of remittance. In IOP Conference Series: Earth and Environmental Science; 2020; Atlanta, GA, US. p. 1-6.

Ajaero

, Onokala

. The Effects of Rural-Urban Migration on Rural Communities of Southeastern Nigeria, International Journal of Population Research 2013;2013:1–10. DOI: 10.1155/2013/610193.

Joly

. Some structural effects of migration on receiving and sending countries, International Migration 2000;38(5):26–40.

Feleke

, Pankhurst

, Bevan

, Lavers

. Migration and rural-urban linkages in ethiopia and implications for policy and development. In Irish Aid–Ethiopia; 2006; University of Bath: Research Group onWellbeing in Developing Countries, Ethiopia Programme. pp. 1-97.

Ferework

. N. An Exploratory Study of Child Out-Migration in Ethiopia: The Gurage Case. 2007.

Worku

. The Impact of Urban Migration in Village Life: The Gurage case, MA thesis in Social Anthropology, Addis Ababa University Addis Ababa, Ethiopia: CODESTRIA- LIBRARY; 1995.

Yousra

, Julie

. Does Migration Improve Living Standards of Migrant-Sending Households? Evidence from Rural Ethiopia. Migrating out of Poverty Programme. 2016; pp. 1-54.

Kuschminder

, Andersson

, Seigel

. Migration and multidimensional well-being in Ethiopia: investigating the role of migrants destinations, Migration and Development 2018;7(3):1–20. DOI: 10.1080/21632324.2018.1463903.

Adams

. Remittances, household expenditure and investment in Guatemala, World Development.- 2018;38(11):1626–1641.. DOI: 10.1016/j.worlddev.2010.03.003.

10.

Adams

, Cuecuecha

. The Impact of Remittances on Investment and Poverty in Ghana, World Development 2013;50(C):24–40. DOI: 10.1016/j.worlddev.2013.04.009.

11.

Zhang

China’s Poor Regions, Rural–urban migration, poverty, economic reform, and urbanisation. 1st ed. London and New York: RoutledgeCurzon; 2003.

12.

Tanle

, Ogunleye-Adetona

, Arthor

. Rural-urban migration and household livelihood in the Agona West Municipality, Ghana, Journal of Geography and Regional Planning 2020;13(1):1–18. DOI: 10.5897/JGRP2017.0670.

13.

Sanusi

, Denis

, Kumar

, Kispotta

, Bose

, Srivastava

. Estimating multidimensional poverty and identifying poor households in Kano state, International Journal of Multidisciplinary Research and Development 2015;2(12):148–157.

14.

Megbowon

. Multidimensional Poverty Analysis of Urban and Rural Households in South Africa, STUDIA UNIVERSITATIS 2018;63(1):3–19. DOI: 10.2478/subboec-2018-0001.

15.

Dehury

, Mohanty

. Multidimensional poverty, household environment and short-term morbidity in India, Genus 2017;73(3):1–23.

16.

Wang

, Wang

. Multidimensional poverty measure and analysis: a case study from Hechi City, China, SpringerPlus 2016;5(642):1–25. DOI: 10.1186/s40064-016-2192-7.

17.

Sulaimon

. Multidimensional poverty and its determinants: Empirical evidence from Nigeria. MPRA Munich Personal RePEc Archive. 2020; pp. 1-20. DOI: 101842.

18.

Amao

, Ayantoye

, Fanifosi

. An analysis of multidimensional poverty and its determinants in rural Nigeria, Journal of Development and Agricultural Economics 2017;9(11):303–311. DOI: 10.5897/JDAE2017.0857.

19.

Nguyen

. The Analysis of Factors Affecting Multi-dimensional Poverty of the Khmer Ethnic in Tra Vinh Province, Vietnam, International Journal of Economics and Financial Issues 2018;8(4):158–162.

20.

Gebrekidan

, Bizuneh

, Cameron

. Determinants of Multidimensional Poverty Among Rural Households in Northern Ethiopia, Journal of Rural and Community Development 2021;16(1):133–151.

21.

Maity

, Buysse

. Multidimensional poverty and the factors influencing the multidimensional poverty status of Bodos’: a case study of Udalguri district, Bodoland, International Journal of Education Economics and Development 2017;8(4):266–285. DOI: 10.1504/IJEED.2017.088820.

22.

Fonta

, Yameogo

, Tinto

, Huysen

, Natama

, Compaore

, et al. Decomposing multidimensional child poverty and its drivers in the Mouhoun region of Burkina Faso, West Africa, BMC Public Health 2020;20(149):1–17. DOI: 10.1186/s12889-020-8254-3.

23.

Dereje

, Abrham

, Alemu

. Does building the resilience of rural households reduce multidimensional poverty? Analysis of Panel Data in Ethiopia, Scientific African 2021;12:1–19. DOI: 10.1016/j.sciaf.2021.e00788.

24.

Meekaew

, Ayuwat

. Factors influencing livelihood security among the fishing migrant households at places of origin in Thailand, International Journal of Recent Technology and Engineering 2019;8(2S3):1573–1579. DOI: 10.35940/ijrte.B1287.0782S319.

25.

Belay

, Bikila

, Demel

, Muluneh

, Mesfin

. Land use and land cover changes driven by expansion of eucalypt plantations in the Western Gurage Watersheds, Centeral-south Ethiopia, Trees, Forests and People 2021;(5):1–12. DOI: 10.1016/j.tf2021.100087.

26.

Mesfin

, Bamlaku

, Admasu

. Crop producers’ technical efficiency and its determinants in Gurage zone, Ethiopia: A comparative analysis using rural-urban migration as a parameter, Cocent Social Science 2021;7(1):1–14. DOI: 10.1080/23311886.2021.1995996.

27.

Feleke

, Pankhurst

, Bevan

, Lavers

. Migration and Rural-Urban Linkages in Ethiopia: Cases studies of five rural and two urban sites in Addis Ababa, Amhara, Oromia and SNNP Regions and Implications for Policy and Developme. In Irish Aid–Ethiopia; 2006; University of Bath: Research Group on Wellbeing in Developing Countries, Ethiopia Programme. pp. 1-97.

28.

Latif , Lei

, Pathan , Hussain , Khan . Factors affecting diffusion and adoption of information and communication technology among rural users in Khyber Pakhtunkhwa Province, Pakistan, International Journal of Information Technology and Management.- 2018;17(4):349–370. DOI: 10.1504/IJITM.2016.10003630.

29.

Cochran

Sampling Techniques. 2nd ed. New York * London: John Wiley and Sons, Inc.; 1963.

30.

Alkire

, Foster

, Seth

, Santos

, Roche

, Ballón

. Multidimensional Poverty Measurement and Analysis. In OPHI. Working Paper No 46. United Kingdom: OXFORD University Press; 2015.

31.

Alkire

, Foster

. Counting and multidimensional poverty measurement, Journal of Public Economics 2011;95(7-8):476–497. DOI: 10.1016/j.jpubeco.2010.11.006.

32.

Alkire

. Dimensions of human development, World Development, The World Bank, Washington, DC, USA 2002;30(2):181–205.

33.

Alkire

. Why the capabilities approach? Journal of Human Development 2005;6(1):115–135. DOI: 10.1080/146498805200034275.

34.

Alkire

, Santos

. Acute Multidimensional Poverty: A New Index for Developing Countries. United Nations Development Programme // Human Development Research Paper 2010/2011;2010:1-142.

35.

Hosmer

, Lemeshow

. Applied logistic regression: Applied probability and statistics New York: Wiley; 1989.

36.

Gujarati

Basic Econometrics 4th Edition

New York. USA.: McGraw-Hill; 2004.

37.

Eric

, Luke

, David

, Julia

, Brendan

. AProbit Model with Heteroskedasticity. 2018.

38.

Oshio , Kan

. Multidimensional poverty and health: evidence from a nationwide survey in Japan, International Journal for Equity in Health 2014;13(128).

39.

Adams

. Remittances, investment and rural asset accumulation in Pakistan, Economic Development and Cultural Change 1086;47(1):155–173. DIO: 10.1086/452390..

40.

Adams

. Remittances, poverty and investment in Guatemala. In Part 1 Migration and Remittances. Washington, DC: World Bank; 2006. pp. 53-80.

41.

Borjas

. Economic Theory and International Migration, International Migration Review 1989;23(3):457–485.

42.

de Haas

. Migration, remittances and regional development in Southern Morocco, Geo Forum 2006;37(4):565–580. DOI: 10.1016/j.geoforum.2005.11.007.

43.

Russell

. Migrant remittances and development, International Migration 1992;30(3/4):267–288.

44.

Lipton

. Migration from the rural areas of poor countries: The impact on rural productivity and income distribution, World Development 1980;8(1):1–10.

45.

Lekha

. Nature of Multidimensional Poverty Incidence in Rural Nepal: Empirical Evidences from Bhalam VDC, Kaski, Economic Literature 2014;12:16–25. DOI: 10.3126/el.v12i0.14882.

46.

Muhammad

, Derick

, Mukesh

, Wilson

, Dipak

, Santosh

. Estimating multidimensional poverty and identifying poor households in Kano state. International Journal of Multidisciplinary Research and Development. pp. 148-157.

47.

Upton

. The Role of Livestock in Economic Development and Poverty Reduction. PPLPIWorking Paper No. 10, Pro-poor Livestock Policy Initiative (PPLPI). Rome; 2004. 841.

48.

Moll

. Costs and benefits of livestock systems and the role of market and nonmarket relationships, Agricultural Economics 2005;32(2):181–193. DOI: 10.1111/j.0169-5150.2005.00210.x.

49.

Birthal

, Singh

. Structure of Rural Income Inequality: A Study in Western Uttar Pradesh, Indian Journal of Agricultural Economics 1995;30(2):168–175.

50.

Thornton

, Kruska

, Henninger

, Kristjanson

, Reid

, Atieno

, et al. Mapping poverty and livestock in the developing world. Nairobi, Kenya; 2002.

51.

Birthal

, Ali

. Potential of livestock sector in rural transformation in India: The Role of Non-farm Sector (Rohini Nayyar andANSharma editors) Institute for Human Development New Delhi: Manohar Publishers and Distributors; 2005.

52.

Minot

, Epprecht

, Tram Anh

, Trung

. Income Diversification and Poverty in the Northern Uplands of Vietnam. Research Report Abstract 145. Washington DC: International Food Policy Research Institute; 2006. Report No.: 10.2499/0896291480.