The Determinates of Income Inequality in Urban Ethiopia: The Case of Woldia Town

This study examines the determinants of income inequality in woldia town, one of the zonal town in Amhara region in Ethiopia.. For the successful accomplishment of the study, primary data obtained from surveying the households of the town is applied. The inequality situation in this town is analyzed using both Lorenz curve and gini coefficient and income distribution is proved to be highly unequal even higher than the national average with a Lorenz curve far away from the equality line and the gini coefficient of 0.39. In addition to this, the OLS estimation coefficient declared the existence of direct positive effect of level of education on income but inverse relationship between income and dependency ratio. Moreover income of male headed households is greater than that of female headed and those household heads hired in public sectors earn income less than the private sector employees.

, attempts at summarizing the effects of technological change on the relative shares of income in the neoclassical framework. His conclusions indicate that the inclusion of technological progress is certainly a welcome development, but that this did not lead to a major leap forward in economic research: Simon(2013),conduct a research paper on the main factors behind high level of income inequality in sub-Saharan African countries using up-to-date panel data set from countries for the period between 1990 and 2010. Based on the results from random effect regression, variables such as government expenditure, the level of education and the existence of democracy are investigated to be important variables in reducing income inequality. However, foreign aid is found to increase the level of income dispersion since it does not benefits the poor households. Abebe (2016), conducting a study using general entropy index with the aim of analyzing the determinants income inequality among sampled households who find themselves at the bottom and top of the income or consumption distribution in urban centers of South Wollo administrative Zone of Ethiopia. Findings of this paper suggests that widening access to education, supporting informal sector, urban agriculture and creation of job opportunities, urban investment to improve access to urban land urban infrastructure, the quality of life and housing development have a significant impact on reducing household's income variation Tassew (2009), in his study on poverty and inequality analysis in Ethiopia, he found that even if income inequality remained unchanged in rural areas, there was a substantial increase in urban areas income dispersion. More over the results in this paper show, in Ethiopia, income growth reduces poverty and increases in inequality increase poverty; the income-poverty elasticity lies in the range of -1.7 to -2.2. In rural Ethiopia, the increase in consumption has led to a reduction in headcount poverty.
Okidi (2004) wrote a paper entitled "Understanding the determinants of income inequality in Uganda." Their article is interesting because Uganda in the last period of ten years experienced gradual and sustained growth and poverty decline. Benefits of growth, however, are not being distributed equally. This study provides insights into deepening understanding of the determinants of income inequality in Uganda. Decompositions by subgroups revealed that household characteristics are influential components of overall inequality, a finding also supported by the results based on the regression analysis. Eskindir (2011) shows the important effect of income inequality in poverty reduction using a household level data collected from Bench-Maji zone, SNNP of south west Ethiopia. With the aim of investigating the determinants of income inequality using in equality decomposition analysis approach, uses a data collected from 120 sampled rural households who live in sheko district of this zone. Finally, the result of this paper indicates that the Gini coefficient of the study area is 0.39, which shows that the income distribution in the study area is inequitable. The relative contribution each sources of income to the overall income inequality indicated as: crop production 0.35, livestock 0.01 and nonfarm incomes 0.03. The result shows that much of the income disparity is attributed to income generated from crop production. It was found out that the other income sources have an inequality decreasing effect, that is a raise in income from non-farm income and livestock is favorable for income distribution. Land holding, land allocated for perennial crops and livestock are household variables which have higher inequality weight. Increase in education and livestock variables reduces the income gap whereas land holding, land allocated for perennial crops & annual crops, and household size widen the gap. Concerned institutions in improving rural equity should give high attention on nonfarm income generating activities, and improving the productivity of livestock.

Methodology 3.1. Description of the study area
Woldia is the capital town of North Wollo zone which is one of the eleven zones of Amhara region of Ethiopia, which was established around 1788 by Ras Ali the Great and has serve as the administrative and urban center. It is a nodal town connecting four main roads from opposite directions. It is found 520 km north of Addis Ababa, 320 km east of the Amhara National Regional State city of Bahir Dar, 260 km south of Mekelle, 720 km west of Djibouti. It is situated at 2000 meter above sea level with 350 mm average annual rainfall and 22 o c average daily temperatures (Woldia City Administration (WCA), 2007). The town has hill topography and surrounded by mountains. Although the surrounding areas of the town were covered by forests, nowadays it is dry and stony due to successive deforestation.

Data types, sources and analysis
The analysis in this paper is mainly based on total monthly household disposable income, defined as the sum of incomes from wages and salaries, formal businesses, female household businesses, children's activities, pensions and remittances, farming and livestock.
The type of data used in this study is a primary data collected from randomly selected 220 sample households who are living in woldia town. Using this data both descriptive and econometric methods of data analysis are applied to study the determinants of income inequality. The Lorenz Curve and Gini coefficient measure of income inequality are applied to determine the extent of income dispersion in the town and the significance and impact of variables is analyzed through applying the simple OLS estimation coefficients.

Sample size determination
Selecting too large sample size not recommended due to high cost and shortage of time. Also selecting too little sample size main not truly represent the population character and leads to biased result. In order to determine a representative sample size from selected kebeles, in this paper, finite population correction factor (FPCF) formula developed by Morgan and Krejcie in 1970 is applied. n = z pqN N − 1 e + Z pq Where, n = required sample size N = total population Z = z -score (e.g., 95% = 2.005) e = margin error (rate of accuracy) p = proportion of population q = 1-p To determine the proportion of the population, this study applied a pilot survey. 30 pilot sample households where selected and asked either they have been negatively affected by inequality or not. Out of the pilot sample 21 households answered as they are negatively affected by the existed income inequality. Then, p = 0.7 and = 0.3. Using 95% confident interval, 5% margin error, total population N= 3428 the sample size is determined as follows.2894, 12.34 n = 2.005 0.7 0.3 3428 3427 0.05 + 2.005 0.7 0.3 == 220

Data analysis results and discussions 4.1. Descriptive analysis
Under this method of data analysis different methods of data analysis such as ratios, figures, percentages, means, variance, standard deviations and graphs are used to analyze the characteristics of the data set. Among these descriptive analyses, more emphasis is given to graphical method to measure Gini coefficients of both income and consumption among households. 81.8% 18.2% The above table shows a descriptive statistics on the categorical dummy variables used as explanatory variables included in the model. According to results on the data set 166 out of 220 sampled household heads are male which accounts for 75.4% of the total sample size. The remaining 54 households are managed and administered by female heads and the numeric figure for female headed households is about 24.6%. With regard to the marital status, 75.9% of the households are married and the remaining 24.1% are living separately due to divorcing, widowed or not married. In addition to this 87.7% of the population is estimated to be engaged at work and 12.2% are retired from work either because of old age or being disable at the time of military workers. When we consider the sectors that household heads are hired, 68.6% are engaging in the private sectors out of which 18.2 % are earning their income from underground economies of contraband and street vending activities. The remaining 31.4% are working on the public sectors of the town. and maximum ages respectively. The age of households vary with the variance and standard deviations of 120.4 and 10.97 numerical figures respectively. The household's level of education is ordered starting from the illiterates to those who have a doctorate degree. The average education level of households is either a preparatory or first degree stage.

Descriptive statistics of continues variables
The family size of the households ranges from 1 to 11with the average household members of five individuals. The family size varies from households to households with the variance of 3.13 out of which 24% of the populations have individuals more than the average numbers of 5members.
Summery on the dependency ratio of the town shows that on average one individual is dependent on two working individuals as measured by a mean dependency ratio of 60.4 even if the dependency ration varies from zero to 400 with a variance and standard deviations of 50.58 and 7.01

Summery statistics on distribution of income Description on distribution of income
The distribution of income can be analyzed using descriptive summery measure such as mean, standard deviation quartile, percentile and decile ratios in addition to the Lorenz curve and gini indexes. In this part percentile ratio is applied as a major tool of distribution analysis. The based on table presented below distribution of income is summarized by using the percentile ratio of income in Woldia town.
Based on the results of the data form households, 50% of the individual households of woldia town earns an average monthly income of 400 as indicated by the fifty percentile ratio. The mean value of income is 5190 even if individual household's absolute income varies from 550 to 70000 with the standard deviation of 6900.
The income distribution is skewed to the right by 6.99%. This indicates that only six up to seven percent of the population is earning income greater than the mean value. The kurtosis for income measures a value of 62.2 which is greater than three by large amount to indicate the presence of presence of leptokurtic distribution which measures more out layers than the normal distribution. According to the table 4.1 above, the poorest 1%, 5%, 10% and 25% percent of the population is earning an average monthly income of 600,770,810 and 2500 respectively. This means that the total income share of the bottom 25% of the population is only 6.4% of the total income.
In opposite to this small proportion of the top income earners are earning highest proportion of the total income. The richest 1%, 5%, 10% and 25% of the total population are earning average income of 25000, 10000, 8000 and 6000 respectively and this shows that there is a high level of income inequality in woldia town where few top income earners are exploiting the income of majority poor (top 25% rich households are earning 54% of the total income).
In addition to the above distribution measurements, Lorenz curve and gini index measures of inequality are applied which do not depend on the mean of the distribution instead, inequality is concerned with distribution. In order to measure income and consumption expenditure of households' inequality in the study area, the Lorenz curve and Gini index/coefficient are used.

The Lorenz curve and gini index of income
The Lorenz curve is one measure of income inequality through indicating by how much amount the distribution is far away from the equality line. Any distribution of income with a Lorenz curve near to the equality line represents relatively equal income distribution and if the Lorenz curve for a given distribution is far away the line of equality the distribution is highly unequal. As indicate in the Lorenz curve graph (fig 4.1) the distribution of income in this study area is high as indicated by the down ward bending curve.

Figure 2: Lorenz curve for income distribution in Woldia.
Even if Lorenz can serve as a measure of inequality, it can't indicate the exact quantitative value of the distribution's dispersion. So gini coefficient is the best measure of inequality with the exact number to indicate the level of inequality. It always measures a value between zero and one (between 0 and 100 when calculated as percentage). Gini index is zero when there is equal distribution indicating all individuals under consideration are earning equal income level and it is one in special case when one individual is earning all the income while others are earning nothing.
To derive the value of gini coefficient the excel method of calculating income inequality at households level with the following formula (American Statistical Association, 2014) is applied in this study. = ∑ Where; i = individual household μ = mean value of income n = total sample size xi = income of household i Using this formula the gini coefficient of woldia town is estimated to be 0.439 to indicate the presence of high level of income inequality. So this gini figure of .44 is greater than the 0.33 national average gini coefficient of Ethiopia as measured by World Bank (WB, 2015). The reason for this result is, there is high level of income inequality in urban areas of Ethiopia and relatively low level of income inequality in rural counter parts due to annually earned equal agricultural income. So, high level of income inequality in urban areas will exist when compared with the national average since the average is taken from low inequality rural areas as well.   Vol.10, No.13, 2019 on the model. The calculated F value of 7.11 at 95% critical value indicates that null hypothesis (H0 : all coefficients are zero) is rejected since p value 0.000 which is less than 5% to reject the model's incorrect specification. The R2 value of 0.91 indicates thatvout of the total change in the dependent variable o91% is jointly explained with in the model.

The OLS estimation results
According to the OLS estimation result of the data set, some of the continuous variables have a positive relationship with household's income level and some others have inverse relationship. As the level of education and family size increase by one unit, household's monthly income increases 18% and 6% respectively. The level of education is the significance determinant of income inequality in this town as the level of income will increases by large amount as a level of education increases by one unit. This result is supported by those individuals with low level of education are investigated to be low income earners in United States and in general countries that provide higher-quality education across the economic spectrum, there is much less income disparity (WB, 2013).
On the other hand, dependency ratio does in opposite direction with monthly income. The reason behind it is high number of dependent members with in a household contribute nothing and consuming the income of few independent individuals. In Woldia town, the monthly income of household's monthly income decreases by 0.13 when dependency ratio increases by one percent.
With regards to the effects of dummy variables included in the model, more male headed households earn higher income than the female headed once and the figure shows that the income of male headed is greater than that of female headed by 24%. Furthermore, private sector employees earn more income than the public employees by the respective probability of 23.8%.

Conclusion and policy recommendation
Under this study, determinants of income inequality had been identified. To do so, the widely used measures of income inequality like the Lorenz curve and gini index are applied. The relatively more concave Lorenz curve of income distribution to the origin is obtained to indicate the presence of relatively high income inequality. This distribution is summarized using a quantitative value indicator inequality measure of gini coefficient and the gini index is given to be 0.439. This town's gini index is greater than the national average index of 0.33 because of high level of income inequality in urban areas than rural areas of the country.
In addition to this, The OLS estimation coefficients declared the existence of direct positive effect of level of education on income level but inverse relationship between income and dependency ratio. Income will increase by 17.6% due to a unit change in education level and it will decrease by 0.13% when the dependency ratio increases by one unit. The dummy variables coefficients also shows that the income of male headed households is greater than that of female headed by 24.3% and those household heads hired in public sectors earn income less than the private sector employees by large amount of 23.8%.
High level of education accounts a lot for households to get themselves in high income groups. So household heads of Woldia town has to give emphasis for education, to spend more for schooling for their children and themselves and the government is also required to increase its expenditure on education. When the income of private sector employees is compared with that of public employees, private sectors are paying higher income for their workers. To reduce the level of inequality the governmental public sectors are recommended to pay more than what they are paying now and giving insurance in different forms like health insurance and pension. Additionally, there is male-female income difference and females are earning less than male. To make the balance, gender offices governmental and non-government institutions will give more emphasis for females through training and self-confidence creating activities.