Rural Households Livelihood Portfolios and Determinates of Livelihood Diversication in North-Western Ethiopia

Background : Regardless of the persistent image of rural areas in Ethiopia as a continent of subsistence farmers, over the past decades, there had been an outstanding tendency of rural economic diversification. Numerous motives prompt households and individuals to expand the range of assets, incomes, and activities. This paper is devoted to characterize rural households’ livelihood portfolios and examine the determinants of income diversification using primary data collected from two agro-ecological zones in north-western Ethiopia. To analyze, the data both descriptive and inferential statistics were used. Levels of household livelihood diversification were measured using Simpson Diversification Index (SDI). Censured regression models were employed to identify determinant factors affecting livelihood diversification. Result : The result confirmed that households in the study area collected a significant portion of their income from the diverse farm and off-farm sources. Diversification into off-farm sources contributed 35% to total household income. The result confirms that factors linked to household livelihood diversification measured in Simpsons Diversification Index (SDI) are significantly determined by household head educational status, access for tanning, age of household head, family size, livestock ownership, ox ownership, land owned, the proportion of infertile land, access for road and agro-ecologies. Conclusion : From these results, therefore due attention should be given to strengthening the role plaid by off-farm income in a rural area to facilitate the countries goal of a transformation. Therefore, policy measures need to be directed towards creating conducive conditions through the provision of education and tanning and improve households’ access to credit and improve access to a road.


BACKGROUND OF THE STUDY
In Ethiopia, rural households depend on agriculture and other non-agricultural strategies for their livelihoods. Diversification of income sources (the allocation of productive assets among different income-generating activities) has been put forward as one of the strategies households employ to minimize household income variability and to ensure a minimum level of income.
Bryceson (2002)  incomes, consumption, expenditure, and household food security (Haggblade et al., 2010). The share of off-farm income to the total income ranges from 30% to 50% (Ghimire et al., 2014 andLosch et al., 2012). This process of diversification in SSA has been commonly explained by the combinations of push and pulls factors, which determine the level and type of diversification strategy pursued by a given household (Seid 2016).
Despite the persistent image of Ethiopia as a continent of subsistence farmers, over the past decades, there had been an outstanding tendency of rural economic diversification. Rural farm households regulate activities to find new opportunities and to cope with risks. Numerous motives prompt households and individuals to expand the range of assets, incomes, and activities. These motives comprise different push and pull factors like household size, farm landholding, seasonality of agriculture, increasing price of agricultural inputs, risk aversion, and earn more incomes (Mathewos and Nigatu 2016, Yishak 2017, Seid 2016, Belaineh 2002and Fikru 2008.

Subsequently, the diversity of rural livelihood is receiving increased attention in discussions in
Ethiopia for rural poverty reduction (Prowse 2015, Worku 2016, Tsega and Mary 2013, Yishak et al. 2015, Adugna and Wagayehu 2012, Yisihake and Abebe 2015, Mathewos and Nigatu 2016, Yishak 2017, Seid 2016, Yenesew et al. 2015, Woldehanna and Oskam 2001and Brhanu 2016. However, in most of these studies diversification has been measured as either as the amount of income which is derived from off-farm sources (Tsega and Mary, 2013;Davis et al., 2010;Brhanu, 2016) or a number of portfolios (Yishak, 2017;Geremew et al., 2017;Geremew, 2017;Mohammed et al., 2018;Yenesew et al., 2015;Seid, 2016;Tsega and Mary, 2013), which may lead to a wrong conclusion in a case where a household gain most of their income from a single source while the rest only from more than one sources. On top of this, in all of these studies so far on diversification portfolios analyses, the values of plantation income were underestimated. Tree plantation is one of the economically acceptable opportunities of income diversification in most highland areas of Ethiopia and the Amhara regional state (Bekele 2011: BoEPLAU, 2015: Wubalem et al., 2019, Sirawdink et al., 2011Tilashwork et al., 2013).
Alongside the main objective set in Ethiopian rural policy to attain food self-sufficiency by accelerating the transformation of subsistence agriculture to market-oriented agriculture, it has been not able to generate the desperately needed rural transformation. The possible reasons could be the little attention given to diversification to off-farm and non-farm employment in rural areas. For example, the goal set in the Growth and Transformation Plan (GTP) is silent on the role of the rural off-farm sector. One reason for this lack of emphasis could be the unavailability 4 of solid and up-to-date empirical evidence on the role and determinants of income diversification. Hence, from all the above backgrounds, it is essential in this research to evaluate the level of livelihood diversification considering both the type of income source and share of income and factors affecting a level of diversification. Precisely, the goal of this research is to characterize household livelihood portfolios, determine the extent of livelihood diversification and examine factors affecting the level of household income diversification in North-Western Ethiopia.
The result of this research will provide a clear picture for policymakers when planning for agricultural and rural development by identifying possible bottlenecks of a rural transformation.
Further, the results of our study will also help to learn about which off-farm and non-farm economic activities to pay attention to and the infrastructure that will reduce entry barriers and facilitate easier access to these activities.

Data type and data source
The study was conducted in three districts of the Central Gondar zone in Amhara regional State.
Centre Gondar zone was formerly named as North Gondar zone along with current west and north Gondar zones. A combination of quantitative and qualitative data collected from primary and secondary sources was used. Following the livelihoods approach, in this research household was used as the unit of analysis for the sample survey, as it is considered a suitable unit of analysis for the study of livelihoods (Ellis, 2000). Thus, in the research list of households in each sample kebele's were taken to form a sample frame to select sample through random sampling techniques. Sample sizes of the study were determined by the Cochran formula. It was used for 5 its potential to allow calculating an ideal sample size given a desired level of precision, desired confidence level, and the estimated proportion of the attribute present in the population (Cochran, 1977). The formula is Where; e is the desired level of precision (i.e. the margin of error), p is the (estimated) proportion of the population that has the attribute in question, and q is 1p. There are several ways to measure livelihood diversification. For this research to measure the level of household livelihood diversification Simpson's Index of Diversity (SID) were used because SID takes into consideration both the number of income sources as well how evenly the distributions of the income between the different sources (Minot et al., 2006 andJianmei andPeter, 2013). Besides, in this study Simpson index of diversity is used because of its computational simplicity, robustness, and wider applicability (Jianmei and Peter, 2013). The formula for the Simpson index is given below: Where SID is Simpson's index of diversity, N is the total number of income sources (including forest income) and Pi represents the income proportion of i-th income sources including farm 6 income and off-farm incomes, which are classified based on different empirical works of literature explained in the literature section. Its value lies between zero and one. The Simpson index of diversity is affected both by the number of income sources as well as by the distribution of income among different sources. The SID model is expressed in this study as: Where: Y is Total Household Income, y is an income coming from all sources i i=1, 2, 3, 4….7, farm and off-farm incomes.
Furthermore, to examine factors affecting livelihood diversification Tobit regressions were estimated. As described above the value of livelihood diversification measured in Simpson Diversification Index (SDI) ranges between zero and 1. Thus, conventional linear regression methods have difficulties in explaining the qualitative difference between these zeroes and continuous observations. Tobit model is more suitable to find the parameter estimates if latent or censored sample presents in the dependent variable (Gujarati, 2004). It is specified as follows: Where, Y is the value of SDI, βo is the constant term, β n is parameters to be estimated, X is a set of household characteristics and ε is the error term. For different values of independent variables the equation to evaluate factors affecting household livelihood diversification is becomes: . The study estimated this model using the Maximum Likelihood (ML) procedure. livelihood security (Alionovi et al. 2010). Accordingly, sample households' livelihoods were grouped into sex-sub categories of livelihood portfolios. These sectors, income sources were identified based on empirical literature and FGD discussion. These categories are shown in  Almost all households (over 95.6%) in the study area were involved in crop farming. Besides a high level of engagement in crop production, households in the study area participate in different non-farm and wage farm activities. In the study, it was found that about 85% of households have access to some income from activities other than agriculture (crop production and livestock).
From this 62.6 %, 56.1%, 16.4%, and 50.4% of respondents had reported that the participant and collect some portion of their total income from non-farm employment, plantation, and agricultural waged employment and from non-labor income, respectively (see Table 2:1). It is also indicated that the main source of income in the area, on-farm activity comprises about 65% of the total income on average (see Figure 2-1). The rest 35 % of households' income is from other engagement, where non-farm employment, agricultural wages employment, and non-labor income contributed 23.2%, 5.47%, and 6.6% of the total income, respectively (see Table 2:1). The finding of the study thus confirms the notion of heterogeneous livelihood activities pursued by rural households in the study area. This result is more or less similar to the finding by Adugna and Wagayehu (2012) in southern Ethiopia who reported that non-farm income constitutes 22.8% of the total income. Similar findings were also found by Dereje and Desale (2016) in his study of rural livelihood strategies and household food security in Oromia regional states. He noted that 37.2 % of the respondents supplement agricultural livelihood activities with non-farm livelihood strategies. Others undertake the non-farm as their main livelihood activities. Beyene (2008) also found that more than half of rural households have one or more members participating in activities outside agriculture.

Description of Livelihood Strategies in the Study Area
This section of the research deals with characteristics of diverse livelihood portfolios in the study area.

On-Farm Livelihood Strategies
The livelihoods of the surveyed households were mainly dependent on agricultural activities. In the study area majority of households undertook mixed farming activities, involving both crop production and animal husbandry.

11
Besides, different types of annual crops and long-term crops (trees) are grown in the study area.
In the study area tree crop is one of the dominant long-term cash crops planted on the farm. It accounts for 12.53% of household total income and 19.38% of household on-farm income. Tree plantation is generally viewed as a shift from traditionally grown less remunerative crops to more remunerative crops. It is categorized as secondary agricultural activities (Eurostar, 2013, and Amanor-Boadu, 2013). As described in Figure 1 4 significant portion of the household cultivate a tree as one form of the cash crop. About 55.1 % of households in the study area cultivate both long-term and annual crops. It is also indicated that households' mixing production of trees and other crops is high in highland areas (82%) than midland areas (34.4). Similar findings were also reviled by Abebe (2019), Duguma (2013), and Hailemicael (2012). According to these studies in Ethiopia production and sale of trees allows households to earn more than could be done by allocating the same resources to cereal food production. An increase in the demand of forest product such as wood for construction poles, timber, firewood, charcoal, fencing, posts, farm implements and source of income which makes Eucalyptus is the popular tree crops in the area.
Out of many kinds of eucalyptus species, in the region E. Globulus and E. Camaldulensis are the most widespread of all. The two Eucalyptus species are normally altitude-based with E.
Camaldulensis is being adaptable in the upper midland agro-ecology that is lower altitudes while E. Globulus is mostly found in highland agroecology meaning higher altitudes.
12 The non-farm livelihood category consists of households whose main living is based on activities outside agriculture. These include wage labor in the non-agricultural sector, self-employment in own business, trade of grains, traditional brewing, and livestock. In addition to this, this cluster comprises households who derive their living from formal employment. It comprises two income sources or activities: Non-Farm Self-Employment (NFSE) and Non-Farm Wage-Employment (NFWE). As can be seen from Figure 2-3, 84 % of non-farm incomes were from Non-Farm Self- 14 were struggling to survive by share-cropping out the land we have for two years. However, the output we got from the land was not enough to cover our basic expenses. Therefore, for the last three years, I started traditional brewing under the constraint of labor to work in agriculture. I prepare Tela and Araki. By so doing, I cover my child's school and food for the family''

Agricultural Wage Employment (AWE)
Agricultural wage employment refers to agricultural-related activities which involve the supply of paid labor on farms other than those owned by household members. This includes in the study area contract weeding, milking, crop harvesting, contract farming, and keeping livestock among others. In contrary to involvement in non-farm employment, involvement in agricultural wage employment is generally limited in the study area (only 16.4%) (See Table 2:1). Relatively high levels of involvement in AWE were observed in midland households (18.8%) than highland households (13.2%). However, as indicated in the table the difference in income access from AWE across agroecology is not statistically significant at 5% probability level. This is due to the fact that only the poor and landless participate in this kind of employment. The same table also indicated that AWE income contributes only 5.46 % of the total incomes in the study area.  It was revealed that contract weeding and harvesting are the most common activity in AWE. It is mainly performed by male household members who travel away from their residents to cash crop-producing lowland areas of Metema, Humera, and Kwara. Nevertheless, AWE mainly involves migration, there are also old households and households with many plots in the study area that contract their crop for others to remove weeds, tilling their farmland, and collect harvest based on an agreement of paying in either food or cash. Weeding which involves no migration is mainly done by female household members whereas farming, crop cutting, and collection are mainly carryout by men. Farmers whose plots are located in distant areas also contract their plots for the adjacent farmers for the purpose of weeding and protection.

Non-Labor Income (NLI)
Respondents in this study did not rely only on their labor but also get income from remittance, pension, and informal social support. Non-labor income includes all income as a gift, remittance, donation, aid, other transfer, and compensation. In the study area, as depicted in Table 2:1, 50.4% of households have gathered some income from NLI and the income on average contributes 6.64% of the total income. This income includes income from remittance, pension, and income from renting animals, land, and house. In Table 2:3 it is indicated that from the above-mentioned non-labor income remittance which involves 28% of households and generates a mean income of 2641.8 ETB is the main source. On the other hand, only 10.4% and 0.5% of households gather income from renting animals, land, or house and pension. On average households in the study area collect 897.6 ETB and 98.7 ETB from renting and pension respectively. The Chi-square test indicates that more households in the highland area have had access to non-labor income than midland households and the difference was significant at a 5% 16 probability level (X 2 = 20.844, sig. = 0.000). From FGD participants it was revealed that some households in the highland areas own houses in the nearby town and collect rent incomes. Besides measuring livelihood diversification as either as the amount of income which is derived from off-farm sources or on-farm sources, in this study household level of diversification were assessed following Simpson Diversification Index (SDI). Then, following the approach followed by Ahmed and Melesse (2018) in his study of livelihood diversification, in this study household level of livelihood diversification was classified into four categories based on the standard deviation. Thus, households were categorized as no diversification, low, medium, and high levels of diversification as described in Figure 2-

Figure 2-4 Households level of livelihood diversification
Additionally, Figure 2-5 shows the distribution of the number of livelihood strategies that households in the study area were engaged in during the survey year. Most (28.3.8%) of the households were involved in four livelihood strategies. Another 27.5%, 18.4%, 11.9%, 8%, 4.9%, 0.78% was involved in five, three, two, six, one and seven strategies, respectively. Chi-Square test i.e., χ2 (14) = 217.62; Prob > χ2 = 0.000) revealed that at least one of the predictors' regression coefficient is not equal to zero.

Discussion on Factors Influencing the Extent of Rural Household's Livelihood Portfolio Diversification
Household-level of income diversification and the shares of incomes from the different farm and non-farm activities are presented in Table 2:1. Both farm and non-farm activities are important sources of income for rural households in the sample. As indicated in Table 2:1, almost all households (over 95.6%) in the study area were involved in crop farming and it is the major contributor of the total household income. Diversification into off-farm sources contributed 35% to total household income. Moreover, based on SDI measurement of the level of livelihood diversification, only 5% of the population gathers 100 % of their income from only one source (See Figure 2-4). Hence, zero levels of diversification.
The results of the Tobit regression analysis conducted to estimate the determinants of income diversification are presented in Table 2:4. The results indicate that extent of household livelihood diversification is determined by a number of socio-demographic, economic, institutional, and ecological factors. As reviled in the model result, it was found that some factors had a positive effect on income diversification, whereas others had a negative effect. The results are summarized in Table 2:4 and the subsequent section comprehends discussion of significant variables.

Socio-Demographic Factors
The sex of household heads has a negative but insignificant relationship with the household level of diversification. The negative sign of this variable indicates that households headed by females are less likely to diversify their income sources. This finding (the negative sign) is supported by the findings of Fufa (2015), Amanze et al., (2015), and Adepoju and Oyewole (2014). It implies that male-headed households were able to participate in more livelihood portfolios and collect the diversified income.
Perhaps this may be because as observed in the study area there is a traditional culture that leads to gender disparity which creates female-headed households having less information and chance to join in another form of income sources other than the usual agricultural activities.
Unlike the researcher's expectation, the age of household head positively influences the level of household income diversification at less than 1% probability level. This implies that as the age of household head increases, household's levels of engagement in diverse income portfolio also increases.
Conceivably this could be due to the fact that aged household heads have more adult family members who might be engaged in various livelihood portfolios. Hence, those adult family members can engage in off-farm employment to support their family. This finding is consistent with the findings of Vinefall (2015), Akpan et al. (2016);Gecho (2016), and Irohibe and Agwu (2014). However, it is inconsistent with the researcher's expectation and money other similar researchers finding. These include the finding of Senadza (2012), Fufa (2015), Agyeman (2014), and Amanze et al. (2015) in Ethiopia. According to these researchers finding as the farmers' age increases their risk aversion behavior also increases. As a result older farmers hesitate to invest their money in a new business to diversify their income.
Households' family size measured in adult equivalent is another demographic factor. It has positive significant effect on income diversification at a 5% significance level. In another word, it indicates that as the number of family members measured in adult equivalent increases, the probability of the household to earn income from diversified sources increases. In a rural area, more family means more labor which able households to engage in diverse livelihood activities. Thus, in the Tobit result, it is indicated that an extra member increase in family size measured in Adult Equivalent (AE) would increase the household level of livelihood diversification by 0.018. This finding is consistent with the finding by Adpoju and Oyewole (2014); Ghimire et al. (2014), Idowu et al. (2011), andAgata et al. (2009). Perhaps it could be due to scarcity of land. When there is a large family size, as land is a fixed input there would be high numbers of under-employed family members whose marginal productivity is zero. Therefore, such households might try to find other alternatives employments and diversify their income sources.
In the study it is also found that, households headed by educated household heads had a positive and significant effect on the level of income diversification as education increases the household's opportunity of livelihood portfolios by providing the necessary skills. It also promotes job mobility and skill acquisition that could be needed to engage in other economic activities. This implies that being illiterate has a negative significant effect on income diversification. Education increases households' motivations to obtain income from self-employment and wage-employment activities in the non-farm sector. Thus, from the finding indicated by  (2001) Table 2:4, ox ownership of households is inversely associated with livelihood diversification.
Ox ownership in the study area is an indication of access to animal plow since it is the only source of power in plunging. This result is probably because household who owns livestock may 22 not be forced to diversify their income outside agriculture particularly towards agricultural wage employment and other low rewarding non-agricultural employments.
The other most important asset and economic factor in rural areas is farmland. It was found that farm size had a negative and significant influence on the probability of household engagement in income diversification at less than 5% probability level. The possible reason for negative relation could be as farmers' holding increases they may not have extra time or labor to invest outside agricultural income sources as more time and labor is required to cultivate their land. In addition, households will not be forced to diversify their income sources through various means because they are likely to produce fairly enough food from their land. Similar findings were also revealed by Tekle (2019) Access to basic infrastructure and institutions has its own influence on household-level of livelihood diversification. The finding depicted in Table 2:4 indicated that household distance to all-weather roads is negatively associated with a household rate of livelihood diversification.
This could be due to the fact that household access for road determines a household's movability, opportunities to engage in other income-generating activities outside their own location, and market access. In the area, the common non-farm activities such as daily laborer, petty works, selling of local drinks and handicrafts require access to market and road. Moreover, easy access 23 to transport could also imply proximity to other urban areas or nearby towns which are centers for non-farm and off-farm activities. Thus, access to transport significantly increased the level of income diversification at a 1% level of significance. This finding is consistent with the findings of Asmah (2011) and Winters et al. (2009).
Furthermore, from the result it is indicated that access to credit was positively correlated with income diversification among households in the study area. The more access a household has to formal credit, which is measured by its access to formal credit in the last five years, the higher the household diversifies its income. It is because it relaxes liquidity constraints (Teame, 2018).
However, such an effect is not significant. It could be due to less developed credit access in the study area (only less than 50% of sampled households do have the access). In addition, societies in the locality depend on informal sources of credit, which substitute formal sources of credit.
This result, the positive relationship, is analogous with the findings by Fufa (2015) and Gecho (2016).

Agro-ecology
The study area is characterized by diverse agro-ecologies. It is also one of the determinant factors in the study area. Households in highland agroecology diversify their income source more than households living in midland agroecology at one person significant level. A strong positive relationship was found between being living in a highland area and the extent of diversification.
Households in the highland areas are likely to increase their SID by 0.066 points that as compared to midland agro-ecologies. Perhaps, this could be due to the high proximity of highland households to town and market as well as due to its high access for credit and education as compared to households living in rural areas.

CONCLUSION AND RECOMMENDATION
Despite the persistent image of Ethiopia as a continent of subsistence farmers, over the past decades, there had been an outstanding tendency of rural economic diversification. Only 5% of the population gathers all of their income from a single source. Therefore, zero levels of diversification. However, government policy and interventions so far have been paying attention to agriculture as the only and main sector. But other livelihood diversification strategies have been overlooked in the rural development policies and strategies of the country. Contrary to the policy focus, rural households in Ethiopia collect a significant portion of their income from secondary agricultural activities (tree plantation) and off-farm employment. A significant portion of households is cultivating trees as one form of the cash crop. Censured regression Tobit model exposed that ten explanatory variables were found to be significant determinant factors of the household extent of diversification measured in Simpson Diversification Index (SDI) up to less than 10% probability level. Thus are household head age and educational status; access for tanning, road and credit; livestock, ox and farmland ownership; family size, proportion of infertile land and agro-ecologies. Based on the findings of this study, it is essential to recommend policymakers give due attention to another source of income in rural areas dominantly engaged in agricultural activities. Therefore, initiatives that seek to increase access to education and training, credit, and road need to be strengthened to enhance opportunities for farm households.