Gender Analysis of Income Distribution among Rural Households: The Case of Sodo Zuria Woreda, Wolaita Zone, SNNPR

This study intended to computelevel of income d istribution among male headed and female headed households and identify major determinants of income among male headed and female headed households. This study used data and information collected from of 154 households, of which 94 male headed and 60 female headed. A mult istage sampling technique was used to select the households. The study employs Gini coefficient to estimate income d istribution; and multip le linear and Quaintiles regression to identify determinants of incomelevel among female-headed households and male-headed households.The key finding of the study is that gender was Significant at 1% probability level and had a positive influence on income. The result of this study reveal that income was more evenly distributed among the male headed households than the female-headed counterparts and participation of female headed households in crop production was lessthan male headed households. The results also show that annual income of maleheaded households was higher by 25.4 % than the income of female headed households. Extension visit, technology and off farm income significantly and positivelyinfluenced incomeof the female-headed households. The find ings of this study entail that policy makers should develop the extension system that increases number o f extension visits to female headed farmers. Efforts should be made to empower and initiate female headed households through various programs that improve their technology uptake and build their confidence to involve in other business activities and intensify their income.It is also suggested that the issue of rural financial service receive greater attentionby government and service providing financial institutions.


INTRODUCTION
Most of developing country rural households depend on agriculture as their main source of income and food. In Ethiopia, over 85% of the population live in the rural areas and depend on subsistence agriculture and generate income fro m agriculture (Elizabeth, 2011).About 48% of the agricultural labor force is driven from female family members (FAO, 2011). According to the World Develop ment Indicators of 2006, Ethiopia is one of the least developing countries in the world, with a per capita Gross National Inco me (GNI) of 110.00 USD (CSA, 2009). In the SNNPR; agriculture is also the backbone of the regional econo my; contributing for about 73% o f the regional GDP and more than 90% of the total emp loyment (BOFED, 2005). In addition to the main reproductive and domestic ro les they are ought to play, in study area rural wo men' productive role comes fro m their involvement in direct crop production, livestock rearing, home management activities and marketing of agricultural products and off-farm activ ities. Generally, wo men contribute greatly to food security at household and at national levels. So, imp roving wo men producers' income imp lies a mu ltid imensional contribution to the overall growth of the country. In SodoZuria, considerable gender differences exist in the agriculture tasks performed by men and wo men on the farm and household levels.A lots of role-played by women farmer in agriculture, however, very few o f them o wn or control productive resources (Opio, 2003). Such resources are land, credit, technical services, market outlets , and in formation and education level. They have not received equitable decision making rights with men farmer. Such limited accesses to productive resource and decision making power to wo men farmers for agricultural p roduction creates income inequality between men and women farmers.

Statement of the problem
Notwithstanding women's contribution to food security, wo men farmers are co mmon ly underestimated and ignored in inco me generating activities and trade negotiations processes. They have experienced few concrete benefits and in several cases have even been adversely affected in their liv ing and improvement conditions as result of the implementation of some policies. In fact, there is a general idea among policy makers, politicians, trade officials and negotiators that trade liberalization will reduce poverty equally for men headed farmers and wo men headed farmers but in reality it not true. The problem of lo w agricu ltural productivity and inequality in income distribution exists among male headed and female headed rural HHs. It is also believed that market access will help to increase inco me and improve the conditions of men headed farmers and wo men headed farmers equally but wo men headed households not easily get market access.
In the study area there is still no research conducted on income distribution among rural female headed and male headed households. Therefore this study is aimed at analyzing the socioeconomic characteristics of rural male and female headed

Research Article
Volume 11 Issue No. 02 http:// ijesc.org/ households of Sodo Zuria Woreda. To determine various activities performed by male-headed and female headed households in agricultural activ ities; analyze the level of inco me distribution between the male and female headed farmers; and to determine the factors of inco me between male headed and female headed households .

Objective
The general objective of this study was to conduct gender analysis of income distribution among rural households where as the specific objectives are:  To examine gender roles in crop production among rural households.  To determine level of inco me distribution between male headed and female headed households.  To identify the major determinants of income among male headed and female headed households.

DATA AND METHODOLOGY Description of the Study Area
Sodo Zuria Woreda is located in Wolaita Zone of South Nation and Nationality

Data source and data requirements
This study used both primary and secondary data. Unit o f analysis for the study was crop producing households in the enset based farming system. The secondary data on the target areas demography and socio-economic data was collected fro m the Woreda Agricultural Develop ment Offices and Women, Children and Youth affair offices and other relevant information

Sample Size and Method of Sampling
Muilt-stages, clustered, randomized samp ling procedure was used for this study. It involved the selection of kebeles, villages and households. SodoZuria, which was purposively selected for this research, is among enset producing Woredas in Wolaita Zone. Out of 31 kebeles of Sodo Zuria Woreda, three kebeles (BosaKacha, DelboAtwaro and DelboWogene) of the Woreda were also rando mly selected. Accordingly, four v illages were selected from DelboAtwaro and DelboWogene each and three villages were fro m BosaKachakebele. Through this procedure one hundred fifty four (94 male headed and 60 female headed households were randomly selected for this research.

Methods of data collection
Data collect ion was conducted with formal interviews of the randomly selected male headed and female headed households using the pre-tested structured questionnaires. Discussion with the key informants such as Woredas crop production and agricultural marketing experts and Women, Children and Youth affair offices was held to obtain general info rmation on the income distribution among female headed and male headed the role in crop production among male headed and female headed rural households.

Methods of data analysis
Both descriptive statistics and econometric model were used for analyzing the data from the survey

Descriptive statistics
Descriptive analysis such as ratios, percentages, frequency distribution, means, ranges and standard deviations were utilized to examine and describe the socio-economic characteristics of male and female headed households engaged agriculture production and the roles of gender in crop production.

Econometric Analysis Esti mation of income level among male-headed and femaleheaded households
We used multiple (OLS) and quaintile regression methods to estimate the effects of independent variables on household's annual income for rural male headed and female headed households. The standard model is based on the human capital earnings function developed Mincer in 1998.OLS equation for estimation of income level as below: Where: lnINCi, the dependent variable, is the natural logarith m of the annual income fo r M HHs and FHHs observation i, and Xi is victor of independent variables , β is coefficient (β 1) the vector of unidentified parameters which will be estimated using OLS method, and εi is a random error term. εi is assumed to satisfy the common properties of zero mean and constant variance (Su and Heshmati, 2013 Stata software has been employed to run regression model.

RESULTS AND DISCUSSIONS
In this chapter, the results of the study are presented and discussed in detail to address the objectives of the research. The household characteristics include household size, age, education, extension visit, farm size, access to credit, off farm income, gender, and use of technology which are hypothesized to influence the level and distribution of income within and between the MHHs and FHHs.
Multiple linear regression and quantile regression analysis were used to analyze the effects of d ifferent household characteristics on the level of household income. Since the Multip le linear regressions establish only the average relationship between the different household characteristics and household income based on the conditional mean function, it does not provide the full picture of the relationship. It will not be helpful to understand the relationship at different points in the conditional distribution of the household income. The quantile regression will, however, allo w achieve the objective of establishing the relationship between the different household characteristics at the median o r other quantile (e.g., 25 th , 75 th percentile) of the household income. Both methods were applied to identify the determinants of the level and distribution of inco me between FHHs and MHHs in the study area. In addition to the OLS and quantile regression, the Gini coefficient and General Entropy of Thiel's indices of inequality were used for the analysis of inco me inequality within and between FHHs and MHHs.

Roles of gender in Crop Production
Gender ro les refer to the rights, responsibilities, expectations, and relationships associated with men and wo men. Gender division of labor among farming communit ies of study area has been common. Both men and wo men have been playing a significant role in the crop production. This result was consistent with the findings of Adunga (2012). In Table 16 above, the majority of MHHs households participated in planting (82%), weeding (69%) and harvesting 64%) of crop production. In contrast, only a minority of the FHHs participated in these activities . The FHHs participated in planting, weeding and harvesting was 18%, 31%and 36%, respectively.
The result fro m data shows that, for total sampled respondent's average labor forces that participated in crop production were 40.72 mandays per year. Disaggregated by gender, average labor force participation in all crop production among the MHHs and FHHs were 28 mandays and 13 mandays per year, respectively. As a survey result indicates in table 16 above the average force for planting, weeding and harvesting for female farmers were 2.77, 4.85 and 4.68 respectively and the average force for planting, weeding and harvesting for male farmers were 13.97, 11.95 and 9.5 respectively. So, fro m results we concluded that male farmers engaged more in crop production than female farmers.

Econometric Result 4.3.1 Interaction of Variables by Gender, Multicollinearity test and Heteroskedasticity test 4.3.1.1 testing the gender interact action effect
In order to see if the effects of the independent variables on income vary by gender, a standard linear regression was run fo r both the MHHs and FHHs separately. The coefficients of the independent variables were co mpared using t-statistics. Results showed that there were no statistically significant differences in the magnitude of the coefficients of the independent variables between the two regression equations for most of the variables except for two variab les, access to extension and access to credit. This imp lies that only credit and extension have varying effects on income depending on gender. The other variables included in the model have the same effect on income, regardless of the gender of the household head.

Multicollinearity test
Standard test of mult icollinearity was applied to clear and structure the data before conducting any formal regression analysis. The VIF (variance in flat ion factor) indicates how much the variance of the coefficient estimate is being inflated bymult icollinearity. A VIF near to one suggests that there is no mu lticollinearity, while a VIF near 10 might cause concern and shows a serious mult icollinearity effect. This means a commonly g iven rule of thumb is that VIFs of 10 or higher (o r equivalently, tolerances of 0.10 or less) may be reason for concern. Multicollinearity starts becoming a concern when the VIF is over 2.5 and the tolerance is under 0.40. As it is shown in Table 18 the independent variables had no serious mu lticollinearity among each other. All the variables had a VIF near to one indicating there was no serious mult icollinearity effect on the model.

Heteroskedasticity test
In addition to mu lticollinearlity test, heteroskedasticity test was also conducted. Results of the heteroskedasticity test revealed that the null hypothesis of constant variance (hemoskedasticity) was not rejected at the chosen significance level showing that there was constant variance for all the explanatory variables.

Quaintile regression
Quantile regression analysis is used to analyse the effects of different variab les on inco me on at various levels of the inco me distribution. (25th, 50th and 75 th percentile ) for FHHs and MHHs households separately.
The results of the quantile regression presented in Table 20 revealed that variables exhibit ing statistically insignificant differences using the OLS regression were found to be significant at other levels. This justifies the appropriateness of the use of quantile regression. For examp le, education was positively and significantly associated with annual inco me distribution at the 25 th for the MHHs and 75 th for the total sample and M HHs, but the OLS estimates of education not statistically significant. This implies that education was more important for the M HHs at the lowest inco me or poorest level (25 th ) and at the highest level (75 th ) and for total samp le it was important at the highest or richest level (75 th ). Th is is consistent with Biwei(2013) who found that education benefits the richer and poorer households but not the medians. With regard to farm size, the coefficient of farm size increased fro m 25 th quantile to median and positively and significantly associated with inco me d istribution in the total sample. It is also positively and significantly associated with the inco me of the MHHs at median of the inco me distribution. But it was not significantly associated with income at 25 th quantile and 75 th income distribution for MHHs. When the farm size of the MHHs increases by one unit, the income of the MHHs increased by 0.27 percent at mean income distribution level. The finding reveals that farm size benefits the MHHs and total sampled at median level income distribution and total sampled at lowest income level.
As for access to technology, the coefficient of technology is positively and significantly related to income in the case of the FHHs. The corresponding coefficient is 62.7 % at 25th quantile and it increases to 80.8% at the 75th quantile which also reveals that increasing use of technology increases income for the total sample of households and FHHs significantly.
When it comes to off-farm inco me, it was positively and significantly associated at 25 th , 50 th and75 th quintiles in the MHHs. It was not significantly related to income for FHHs at 25 th quantile inco me distribution but positively and significantly related at the median and 75 th quantile. Th is imp lies that offfarm inco me benefits MHHs in the all level and did not benefit poorer but benefited the households at median level and the richest FHHs.
Access to credit did not significantly affect annual inco me o f FHHs and the total samp led households in the OLS regression but results of the quantile regression revealed that it significantly affected the 75 th quantile of the inco me distribution total sampled households at 5% significance level. This shows that access to credit was not important to poorer households and FHHs and benefits the richest households in study area. Table 21, 22 and 23shows thatthe Gin i coefficient of total income for the total sample is 0.43. The FHHs and MHHs have Gin i coefficient of 0.48 and 0.39, respectively.This imp lies that income inequality was higher among the FHHs than among the MHHs. In other words, income is relatively evenly distributed among the MHHs than among the FHHs.

Lorenz Curve for FHHs and MHHs
The Gini coefficient can be graphically represented by different areas of the Loren z curve. Figure 5 belo w d isplays the Loren z curves by gender. When the Lorenz curve is near to the perfect line of equality there is lo w Gini coefficient and when Loren z curve is far fro m line of perfect there is high Gin i coefficient. In the figure below M HH's Loren z curve was near to the perfect line than that of the FHHs, indicat ing that income was evenly distributed among male headed household than female headed households. http:// ijesc.org/ Figure: 5. Lo renz curve for Female headed and male headed households

Decomposition of income inequality by income source
Income was decomposed by its sources to assess the contribution of each income source to overall inco me inequality. Gin i coefficient was decomposed to identify how much of the overall inco me inequality is due to any particular source of income and this can be used to determine whether inequality in an income source serves to increase or decrease overall income inequality.  Table 21show that crop inco meaccounts the largest share of the annual inco me, accounting for 51.2% followed by non-farm income that contributed 37.7% of annual inco me and livestock inco me that contributed the remaining 11.1% for total sampled household. A 1% rise in income fro m crop production decrease income inequality by 0.092% for total sampled household. A 1%rises in income fro m livestock and off farm income increase inco me inequality by 0.028 % and 0.064%, respectively. Lorenz curves by gender http:// ijesc.org/ The share of crop inco me was the largest among both the FHHs (50.9%) and MHHs (51.3%) fo llowed by off farm inco me and livestock that contributed 38.1% and 11% for FHHs and 37.4% and 11.2% for MHHs, respectively see Table 22 and 23. Th is is due to high Rk, wh ich is the correlat ion of a household's rank in the distribution of each income to their rank in total income. A high Rk coefficient suggests that a household's rank in the distribution of the source income is strongly correlated with that household's rank in the distribution of total income. showed that crop income decreases income inequality while non-farm and livestock increase inequality. A 1% rise in income fro m crop production decrease income inequality by 0.06% and 0.121% fo r FHHs and MHHs, respectively. In contrast, a1% rise in inco me fro m o ff-farm source would increase income inequality by 0.04% and 0.084 % for FHHs and MHHs, respectively. For FHHs, off farm income significantly affects annual inco me co mpared to the case with the MHHs. Livestock income was almost equally affecting total income fo r MHHs and FHHs. A 1% rise income fro m livestock production increases income inequality by 0.037% and 0.027%, respectively for MHHs and FHHs.

4.3.4Decomposition of income inequality by factor (sub group)
In addition to decomposition by income source, I have implemented income inequality decomposition by factors such as age, education, farm size and others. Since the Gin i coefficient is not perfectly decomposable, I have used Thiel's general entropy of indices -GE (0) and GE (1). Results are presented in Table below.
Accordingly ,it is widely known that an income inequality exist between FHHs and MHHs. Most of MHHs active in accessto technology and market informat ion that made them to have more inco me than FHHs. However, most of FHHs are correlated with domestic works which was not paid high.
Fro m Table 25 below it can be observed that male headed households had an income share of 61% and female headed households had income share was 38%. The inequality ind ices for female headed household were greater than male headed households. The Gini coefficient fo r female headed households has greater than male headed households; means income was easily distributed in male headed household than female headed households. All indices also indicate that inequality within the groups is a greater problem than inequality experienced between the groups. This leads that there was female headed households more inequality contribute than male headed households.
Education level is one o f the most important social statuses in the community. Thiel's indices of inequality indicate that inequality is more widespread in the group with no formal education for both the MHHs and FHHs .The Gini coefficient is also higher for the groups with no formal education than for the other groups. Further, inequality within the groups is greater than inequality experienced between the groups. The group with no formal education contributes the highest amount of inequality to the total inequality. This result similar with the studies by Mankiw et al (1992) using Slow's model finds a negative relationship between education and inequality.
Deco mposition by age indicates that inequality is maximu m in the groups where household heads are older than 64 years o ld. The Gini coefficient for this group is 0.47 co mpared to 0.46 fo r the group of 45-64 years old and 0.38 for those with belo w 45 years old. This shows that inequality was the highest for oldest household groups. The age group of above 64 years is the highest inequality contributor to the total inequality. http:// ijesc.org/ Inequality decomposition by household size in Table 25 below, households of size greater than 5 members have higher inequality than otherwise. The Gin i coefficient for households with a household size greater than five was 0.44 and for those with a household size o f less than five was 0.41. Therefore, household who have house member greater than 5 contributes high inequality for total inequality. In addition, inequality experienced within the groups is greater than that experienced between. In Table 26 belo w Th iel index deco mposition by farm sizein hectareof the household indicates that inequality is higher fo r households with farm size greater than 0.5 hectare. The Gin i coefficient was also high for the households having greater than 0.5 hectares. This shows farm size greater than 0.5 hectares highest inequality contributor to total inequality. There was higher inequality within groups than between groups. This shows that households who access to technology is more contribute inequality for total inequality. http:// ijesc.org/ Access to agricultural extension services has direct influence on the production and incomes of the farmers. The higher access to the extension service, the more likely that farmers adopt new technology and innovation that increase income of the households. Decomposing total inequality by extension visit in Table 27 belo w shows that there was high Thiel index of inequality among households have no access to extension visit than those who have access to extension. This shows that households who have no access to extension contribute high inequality.
Deco mposition by off farm inco me access indicates that inequality is higher in groups of households who have access to off farm income than those who have no access to off farm income. The index shows that there was high inequality with in groups than between groups. This imp lies inequality was high among the households have access to off farm inco me than those have no access to off farm inco me and off farm inco me was highest inequality contributor in study area.

CONCLUS IONS
This study analysed the gender differential in the level and distribution of income of rural households in Sodo Zuria Woreda of the SNNPR. It also identified the major determinants of the variation in both the level and distribution of income. Both linear and quantile regressions were applied to the study data that came fro m154 households, of who m 94 M HHs and 60 FHHs who were selected using a mult istage sampling technique. Quantile regression was used to capture the nonuniform effects of independent variables at different quantile of the income distribution. The level of income inequality was measured using Gini coefficient. Theil's General Entropy indices of inequality were also applied in the case of the decomposition of the total inequality by factor.
This study applied decomposition analysis using Thiel general entropies to examine the impact of different sources of inco me and factors to explain the variat ion in income inequality. The sources of income for MHHs and FHHs were crop income, nonfarm income and livestock inco me. As the co mmunity in the study area is predominantly an agricultural co mmun ity, crop income p lays a dominant role as an inco me source followed by non-farm inco me and livestock. With respect to inequality, nonfarm inco me and livestock inco me represent an inequality increasing source of inco me wh ile crop inco me represents an inequality decreasing source. Results also indicate that inequality within the groups was greater than inequality between the groups.

RECOMMENDATIONS
In the analysis of the role of gender in crop production, the major finding was that there was large inequality in crop roles among MHHs and FHHs. This suggests that the Woreda wo men, children and youth office together with woreda Agriculture offices should mobilize and in itiate FHHs to participate more in crop production. This could be done, on the one hand, by putting in place a system of incentives in the form of inputs and improved agricultural production and processing technologies. On the other hand, men should also be encouraged to share in domestic tasks by putting in place initiat ives and incentives that entice men to work more in the house. Further, effective gender sensitization programmes may be required. This could be done through non-formal educational activ ities, http:// ijesc.org/ agricultural extension, meetings and mainstreaming gender issues in school curricula at all levels. In addition, info rmal educational activities organized for rural farmers should take note of the heavy domestic workload of females. These activities should also be scheduled at appropriate times to enable many FHHs to participate in crop production and market ing. Imp roving participation of wo men farmers in various areas of extension programmes is the best option for empowering farm wo men for better production that increase income. Therefore, it is recommended that the Woreda agriculture office should be organize and conduct training programmes based on wo men's need, in a manner that wo men are encouraged to attend, taking into consideration timing, duration, location and language; in any training organized fo r farmers