The Distribution of Welfare in Uganda¹

Paul Okiira Okwi and Darlison Kaija

Abstract: This paper examines the distribution of welfare in Uganda in 1997. The data used was obtained from a household survey conducted by the Economic Policy Research Centre (EPRC) in the first quarter of 1997. The analysis of the data focused on the distribution of welfare as measured by household consumption expenditures. It also focused on the attributes of the poor and the very poor households and on the characteristics of their component members. The major findings of the analysis are that the poor are predominantly found in the rural areas, are less educated, have large household sizes, and are primarily agricultural workers. They lack basic services and amenities and have very low levels of expenditure. Inequality levels are high in Uganda as shown by the summary Measures of Inequality. Across regions we also see some divergence in welfare distribution and it is conclusively clear that welfare is unequally distributed in Uganda. The poor are sharing very little of the benefits of growth, while the rich are enjoying the greatest share of the benefits.

1. Introduction

Uganda is classified among the poor countries in the world and in Sub-Saharan Africa. It has a per capita income of about US $250 and this is well below that of other developing countries in the region and the world. Within the country, the problems of poverty and the distribution of welfare have raised much concern. Currently, an impression among some economists is that the situation has persistently worsened compared to the 1960's and to the period prior to that decade. The World Bank (1993) reports that many people in Uganda feel that the economic situation is presently harder than in the pre-independence period and in the 1960's. The economic situation in the pre-independence period and the 1960's can be interpreted from data which shows that employment levels increased by about 32%, private consumption expanded at a rate of 5.6% per year and domestic savings were about 15% of GDP1. Furthermore, the expansion of social services such as health and schools, economic infrastructure such as roads, and the supporting services were significant.

In the early 1970's, there was a reversal of the earlier trends. The factors that exacerbated the poverty problem and to some extent almost led the economy to collapse included, among others, economic mismanagement, war and civil strife. Inappropriate fiscal, monetary and trade policies led to problems of debt, fiscal deficits, stagflation and balance of payment. Among the social indicators, poor health conditions and education facilities greatly fanned the poverty problem. The situation did not change much in the 1980's and early 1990's because the provision of basic needs to the people remained inadequate and inequitable. Generally, welfare levels in Uganda are manifested in a number of ways, ranging from nutrition, housing, health, asset ownership, and education to access to essential services (table 1).

Table 1. Uganda: Major economic and social indicators

SOURCE: World Bank. World Development Report (Washington, D.C.: World Bank, 1993).

This paper analyses the distribution of welfare. Given that Uganda is now one of the fastest growing economies in Sub-Saharan Africa, an understanding of the policies that have led to these high rates of growth together with the distribution of welfare may set an example for other African countries. This study stems from the debate on economic growth versus equity among development economists. Uganda seems to have focused more on economic growth and less on equity. This analysis seeks to explore the levels of welfare distribution in Uganda since 1986, after ten years of positive economic growth.

1.1 The Nature of the Problem

Given the recent developments in the Ugandan economy, the distribution of welfare is of interest both to the policy makers and to the common man. After about two decades of economic decline, Uganda began to experience positive economic growth in 1986. Indeed most published data today portray Uganda as one of the fastest growing economies in Sub-Saharan Africa with an average growth rate of about 6 percent per annum over the past ten years. However, the data little focused on the distributional aspects of this development. Therefore, an analysis of the distribution of welfare in 1997 is important in understanding the welfare effects of Uganda's recent economic progress.

This paper is organised into five different sections. The next section will discuss the theoretical background of using household consumption expenditure as a measure of welfare. Section Three examines the distribution of consumption expenditure and other household features by population categories and regions, ranked according to their welfare levels. The fourth section will analyse welfare distribution using different measures of inequality such as the Gini co-efficient, Lorenz curve and Atkinson measures. The fifth section draws up a poverty profile for Uganda and the last concluding section discusses the policy implications of such a distribution of welfare.

2. Framework: The use of Consumption data as an indicator of welfare

In this study, the social welfare function (i.e., social welfare considered as a function of the utility levels of individuals) is used. It is assumed that households/individuals have the same utility function, which is determined by the household's characteristics such as sex, age and household size. In this study, the "money metric" is used to label the indifference curve which possesses two main features, namely, allowing for the distribution between individuals at different levels of utility, given observable data, and not implying any cardinalisation of the common individual utility function.

The money metric utility (the amount of money required to attain a specified utility level) is used to compare the different household/individual utility curves. This is based on the premise that total consumption, as opposed to total income, represents the welfare levels of individuals. The use of consumption rather than income is also justified because consumption is a better indicator of welfare than income, since income fluctuates over short periods of time while consumption is smoother over time. Also, consumption data is more reliable than income data because it is less sensitive than income to the respondent at the time of collection. Respondents are more willing to discuss their consumption expenditures rather than income. Also income is not easy to obtain and estimate for the self-employed workers. In this paper, total consumption expenditure is used to analyse the distribution of welfare.

Another issue that will be involved in this study is the effect of additional household members, especially the children, on household expenditure. Additional members increase the costs of households but children do not increase these costs as much as adults do. Therefore, children are believed to take a small proportion or consume less of the total expenditures and they are considered as fractions of adults. This analysis is derived from Deaton and Muellbauer, 1980. The adult equivalence scale system allocates weights to different members of the household by age and the scale ranges from zero to one (i.e., one can assume a weight between zero and one). Under this system, children will be given smaller weights than adults. The equivalence scales to be used are adopted from WHO (1985) and are also used by Ravallion and Bidani (1994). They are presented in table 2 below. The sum of weights for each household is used to divide household expenditures to arrive at a measure of welfare. Due to data problems, we shall assume that each individual has the same level of welfare as the household as a whole. All comparisons will be made between individuals and also between households after their welfare has been assigned.

Table 2. Equivalence scales based on measures of nutritional requirements

Age	0-11 months Yrs.	1	2	3-4	5-6	7-9	10-11	12-13	14-15	16-17	18+
Weight	0.27	0.37	0.45	0.52	0.62	0.70	0.73	0.80	0.88	0.95	1.0

SOURCE : WHO (1985).

It should be noted that some aspects of welfare such as health of individuals will not be included in the welfare measurement. This is because of the difficulty in measuring the health status of an individual. The welfare measures used in this study are primarily based on consumption measures of welfare and we may ignore other aspects of living standards that are more difficult to measure. Overall, this drives us to test the hypotheses that: Welfare is equally distributed in Uganda and that there is a strong association between education and household welfare.

3. The Distribution of Consumption Expenditure

This section examines how total consumption is distributed by population deciles. In this categorisation, decile 1 is composed of the poorest 10 percent of the population (as measured by both the adjusted and unadjusted per capita consumption expenditure). The second decile contains the next poorest ten percent, etc., and decile 10 contains the richest ten percent of the population. In this case, two figures of per capita expenditure are presented. The adjusted per capita expenditure figures involve the total household consumption expenditure divided by the number of adult equivalents, while the unadjusted per capita expenditure figures are simply total household expenditure divided by the household size.

From the results in table 3 below, it is evident that the weighting of children (adjustment) gives households higher means per capita compared to the unweighted results. There is a big difference between the adjusted and unadjusted mean expenditure (70,776 shillings compared to 45,108 shillings, respectively). This portrays a significant difference although the distribution of welfare seems less affected by the use of either method. Evidence from table 3 shows that there exists a very high degree of inequality in Uganda - the poorest 50 percent of the population receive only 17-18 percent of total consumption expenditure while the richest 10 percent receive 37-39 percent.

This shows a high degree of welfare inequality. This data is comparable to the results obtained in Côte d'Ívoire in 1986. In Côte d'Ivoire, the poorest 50 percent received about 21 percent of total consumption while the wealthiest 10 percent had 33-35 percent. Evidently, Uganda shows a worse distribution of welfare based on these results. A further discussion of inequality will be presented in Section Four.

Table 3: Distribution of total consumption by welfare decile

Decile	Mean per capita monthly expenditure		Percentage of total monthly expenditure
	Adjusted	Unadjusted	Adjusted	Unadjusted
1	10046.6	6970.9	1.4	1.5
2	18163.5	12309.1	2.6	2.7
3	24272.7	16120.1	3.4	3.6
4	30788.9	20693.4	4.4	4.6
5	39889.9	26651.2	5.6	5.9
6	50315.0	33504.9	7.1	7.4
7	63361.8	41785.8	9.0	9.3
8	82262.1	54040.0	11.6	12.0
9	111139.8	72181.1	15.7	16.0
10	276 587.5	168352.7	39.1	37.3
Uganda	70776.8	45108.4	100.0	100.0
SOURCE: Authors' own computations from the survey data (1997).

Note: The sample size is 993 households. The prices used were market prices (for both home produced and marketed items). The prices were not converted using regional price indices, because so far only one price index is available, and that is for Kampala, and it was considered inadequate. However, since most of the results are consistent with those of previous findings, using the price indices would probably have a smaller change in the rankings of welfare.

3.1 Household Characteristics by Expenditure Groups and Regions:

In this subsection, the different welfare levels of the different population groups within the Uganda population will be discussed. The population is divided into five quintiles as shown in Table 4. The first quintile contains the poorest 20 percent of the population as computed by adjusted and unadjusted per capita consumption expenditure. The next quintile contains the next poorest 20 percent etc. and quintile five contains the richest 20 percent of the Ugandan population. The quintiles are then grouped according to the region in which the household resides, the sex of the household head, occupation of the household head and the highest education level attained by the household head. For each group according to the quintile, we compute the mean per capita expenditure using both adjusted and unadjusted figures. These data are presented in table 4 and 5.

Table 4. Characteristics of households by quintiles

	Breakdown within quintiles (%)						Monthly expenditure
Characteristics	All Uganda	1	2	3	4	5	Adjusted	Unadjusted
Sex of head
Male	80.2	83.7	84.3	82.2	77.3	73.7	65346.7	43804.0
Female	19.8	16.3	15.7	17.8	22.7	26.3	92608.0	54934.7
Region
Northern	20.1	27.0	23.4	17.8	21.2	11.7	51783.9	32703.4
Western	29.9	21.9	28.9	37.1	37.4	24.4	61940.7	40015.6
Central	19.5	14.3	17.3	16.2	19.2	30.2	85142.5	58102.7
Kampala	10.2	1.0	4.6	9.1	10.6	24.9	168430.0	106381.1
Eastern	20.2	35.7	25.9	19.8	11.6	8.8	40093.6	26092.6
Occupation of household head
Farming	26.2	50.0	37.6	26.4	9.6	9.3	37890.8	23920.2
Comm. trade	16.7	4.1	10.2	16.8	25.8	26.3	97545.4	60629.9
Civil servant	25.2	9.2	14.2	25.4	35.0	35.6	95574.8	62228.5
Casual labourer	11.4	21.9	17.3	9.6	3.0	5.4	38464.7	33022.0
Others	20.1	14.8	20.3	21.3	25.8	18.5	78270.4	49572.9
Education of household head
None	7.5	14.8	9.1	6.6	3.0	3.9	44711.8	38490.8
Never completed primary	21.2	35.2	29.4	17.3	14.7	9.8	47199.7	29515.6
Completed primary	15.2	21.4	19.8	15.7	8.6	10.7	48387.2	32169.0
Post-primary	21.4	17.4	21.3	27.9	24.2	16.1	62361.8	41301.5
Post-secondary	3.6	1.5	3.1	3.1	3.5	6.8	98156.8	57212.7
Tertiary level	18.9	6.1	10.2	20.8	27.8	29.3	103347.9	64452.7
University or higher educ.	9.4	2.6	3.6	6.1	15.2	19.0	99692.1	69441.9

SOURCE: Authors' own computation from the data.

According to regions, we see that the Central Region has the largest proportion of the rich households and individuals in Uganda. It has the largest percentage of the population in Quintile 5. This region is the principal producer of Uganda's main cash crop, coffee, and therefore it is not surprising that it is better off than other regions. Kampala, which is taken as a region because of its unique characteristics and the diverse composition of the population, has the highest per capita expenditure when we use both the adjusted and unadjusted per capita figures. It ranks highest, being an urban centre and capital city, so that it is not surprising that expenditure levels are high, given the high cost of living. These high expenditures also reflect the relatively high earnings in Kampala because it is composed mainly of the working class and commercial workers. It therefore requires a careful interpretation of the data since the prices have not been adjusted using the price indices. Leaving Kampala as part of Central Uganda would exaggerate the situation for the Central Region.

The poorest region is the Eastern Region followed by the Northern Region. This result is not consistent with earlier findings by the World Bank (1993 and 1995), which show that the poorest region is the Northern Region. This change in ranking could be attributed to the severe drought and cattle raids experienced in the East during the previous years. The Western Region has the biggest representation in the middle quintiles.

Interestingly, the findings by Kaija (1995) and World Bank (1995) are confirmed in table 4. Households headed by a female show better welfare levels compared to those headed by males. Before adjusting for adult equivalence scales, female-headed households spend shillings 54,934 compared to shillings 43, 804 for the males. Even after adjusting, the result does not change but instead the gap increases to shillings 92,600 compared to 65,346 for male-headed households. This finding contrasts with several earlier studies such as World Bank (1993), which show that households headed by females in general tend to be poorer than those headed by males. It is also consistent with the findings in Côte d'Ivoire. The simple explanation for this could be that today, women in Uganda are more concerned about their welfare than their male counterparts and the increasing awareness among women (mainly through Non-Governmental Organisations) could have contributed to this. Even an examination across regions shows that except for the Western Region, in all the other regions, female-headed households have higher per capita expenditure levels both before and after adjusting for adult equivalence scales.

The distribution of welfare by activity status of the household head shows that working as a civil servant for the government yields higher welfare levels. Given that most civil servants are educated, it is likely that this factor could have contributed to this difference. Households with heads working as farmers have the lowest welfare levels, as they contribute the biggest percentage of the lowest quintile and have the lowest average per capita expenditures (both adjusted and unadjusted), and they contribute about half the proportion of the lowest quintile. These are followed by the casual labourers while the commercial workers constitute the biggest proportion of the middle groups.

Glewwe and Van der Gaag (1990) found that education has a strong correlation with welfare and that it determines the income of workers. From table 4, 15 percent of the poorest quintile lived in households where the household head had no education at all. Going by completion of primary school, about 71 percent of the poorest quintile lived in households where the head had at least primary education. From table 4, one can also see that households with educated heads constitute the largest percentage of the better off (richest 20 percent) population. Even the per capita expenditure levels reveal that households headed by educated heads spend more than the ones with lower education. We also see that the level that constitutes the highest percentage of the rich includes those households with tertiary education. An explanation for this could be that there are quite a few households with heads who have attained university or higher education. The limitation is basically due to infrastructure. Even at the regional level, we can see that households with higher educational levels have better standards of living compared to their less educated counterparts. However, the more educated tend to be concentrated in the urban areas like Kampala, hence the higher welfare levels in these areas.

Table 5. Characteristics of households by region

Characteristics	All Uganda	Northern	Western	Central	Kampala	Eastern
Sex of head
Female	80.2	22.5	12.5	25.3	31.7	16.9
Male	19.8	77.5	87.5	74.7	68.3	83.1
Education of head
None	7.5	8.0	10.1	6.7	3.0	6.0
Never completed P/S	21.2	26.0	14.1	23.7	11.9	28.9
Completed primary	15.2	18.0	14.5	10.3	13.9	18.9
Post-primary	21.4	30.5	16.5	16.5	20.8	24.4
Post-secondary	3.6	2.5	3.7	5.2	6.9	1.5
Tertiary level	18.9	12.0	21.6	25.8	22.8	13.4
University or higher educ.	9.4	2.5	15.8	6.7	18.8	4.5
Occupation of head
Farming	26.2	44.5	12.8	37.6	1.0	29.4
Commercial trade	16.7	12.0	12.5	23.2	26.7	16.4
Civil servant	25.2	19.0	28.6	29.4	19.8	24.9
Casual labourer	11.4	8.0	13.1	5.2	6.9	20.4
Others	20.1	16.5	33.0	3.6	44.6	8.5
Mean per capita expenditure
Adjusted	70776.8	51783.9	61940.7	85142.5	168430.0	40093.6
Unadjusted	46014.4	32703.4	40015.6	58102.7	106381.1	26092.6

SOURCE: Authors' own computation from the data

It is a fact that education significantly determines household welfare as seen from the distribution by quintile and region. This relationship has serious implications for policy in Uganda, and perhaps a strong justification for the current government move towards providing more equitable education to all if the benefits of economic growth are to be more fairly distributed. Recently, the government of Uganda adopted a Universal Primary Education (UPE) policy in an attempt to strengthen the education levels in the household. This policy is strongly supported by the evidence provided in this study and seems a more equitable way to distribute the benefits of economic growth.

In terms of housing characteristics, table 6 provides information on the type of lighting, source of drinking water, type of fuel and building materials by welfare quintiles and region. The analysis of these utilities may seem surprising at first sight but the data suggests that many households are poor along this dimension. The justification of this section is that a household without these basic utilities is prone to disease and suffering.

Access to safe drinking water is one of the major determinants of household welfare. Uganda has a variety of sources of drinking water but it is evident from the table 6 that the provision of this service is extremely poor, especially in the lower or first quintiles. Accordingly, 42.35 percent of the households in the first Quintile do not have access to clean drinking water. About 20 percent use river, lake or spring water while 5 percent use tap water. The poor access to safe water exposes the people to water-borne diseases. A similar story exists at the regional level. The Eastern Region has 38.8 percent of the households without clean drinking water and yet is also the region best served with clean drinking water. It is interesting that Kampala, which is the richest and the capital city, has the greatest percentage of people without clean drinking water. The provision of safe drinking water is vital since it also determines the health conditions of people. It is therefore clear that the policy on provision of safe drinking water may have been neglected.

Table 6. Housing characteristics by quintile

SOURCE: Authors' computation from the data.

Turning to the fuel type, the comparison of "firewood" and other fuel sources may actually seem surprising at first sight. The results on table 6 show that 46 percent of the population use firewood for cooking and the regional level evidence in table 7 also reveals high use of firewood. Of course this has serious implications for the environment. The poorer households use more firewood (75 %) compared to the wealthier households (18%). At the regional level (see table 7), all five regions show very large numbers of poor households using firewood for cooking. The northern region tops the ranking with about 64 percent of the households using firewood. The use of firewood as fuel also has implications for the household's health conditions in terms of clean air, and for the environment in terms of high degradation. The results are even more puzzling when charcoal is included as another form of fuel wood. Charcoal also contributes significantly to the fuel type used. In major urban settlements this is the main type of fuel used. It so far leaves no doubt that the environment is set to be destroyed at unprecedented rates given that the population is also increasing at high rates and reclamation of land is also high. It is startling that only 8 percent of the households use electricity for cooking. This leaves several questions to be answered. Uganda is a major exporter of electricity in the region yet the local supply and use of electricity is very low. Why is this so?

Table 7. Housing characteristics by region

Characteristics	Region
Light	Northern	Western	Central	Kampala	Eastern
Lantern	16.1	18.4	11.9	6.9	9.5
Tadooba	47.7	39.8	49.0	16.8	38.0
Candle wax	1.5	0.7	2.1	9.9	3.5
Firewood	25.6	1.0	1.6	2.0	16.0
Electricity	7.5	38.4	32.5	53.5	6.0
Gas	0.5	1.0	2.6	5.0	0.5
Solar	0.5	0.7	0.0	0.0	0.5
Other	0.5	0.0	0.5	5.9	26.0
Fuel type
Firewood	63.8	35.1	59.6	5.0	51.2
Charcoal	24.6	31.4	24.4	22.8	37.3
Kerosene	10.6	25.7	7.8	32.7	10.0
Electricity	1.0	7.4	7.3	34.7	1.5
Other	0.0	0.3	1.0	5.0	0.0
Drinking water
Piped in dwelling	1.0	15.5	2.1	25.7	0.0
Pip outside dwelling	0.5	11.5	6.2	15.8	0.0
Public tap	0.0	8.4	14.4	43.6	11.4
Borehole	81.5	9.1	2.6	0.0	21.9
Protected spring	8.0	26.9	36.6	10.9	23.9
Unprotected spring	6.5	15.5	31.4	3.0	16.4
Rain water	0.0	2.4	1.6	0.0	2.0
Vendor/truck	0.0	0.0	3.1	0.0	1.5
River or other	2.5	9.4	1.0	0.0	22.4

SOURCE: Authors' own computation from the data.

The lighting conditions show that the traditional form of light in Ugandan households is the tadooba. It is upsetting to see that about 40 percent of the poor households in Uganda use tadooba for lighting. More perilous conditions are exhibited in cases where poor households have to use reeds or firewood for lighting, a common practice in Northern Uganda for families that cannot afford kerosene. This situation depicts a high form of inequality because the wealthier households use more electricity. For health reasons, a tadooba is harmful when used for lighting. It gives off quantities of soot, which is a mixture of many non-hygienic gases. Given that the poor are already vulnerable, then what is the fate of a rural poor person who gets his health damaged because of pollutants that could otherwise be avoided?

4. Welfare Inequality in Uganda

Previous studies, such as World Bank (1995) and Opio (1997), show some degree of income inequality in Uganda. In this section, to analyse welfare inequality, summary measures of expenditure inequality are used. These measures are single numbers that describe the entire distribution of expenditure by region and national level. Both adjusted and unadjusted figures of per capita consumption expenditure are used. The principal measures of inequality used in this study are the Lorenz curve, Gini coefficient, Atkinson's coefficient, the Log of Variance of expenditure and the standard deviation. These methods are reviewed and presented in Deaton (1997) and Anand (1983). However, we provide a brief definition and justification for using some of them. A group decomposition of some of the measures among the regions will be done. This kind of decomposition is justified on the ground that one can determine the potential effect on overall inequality of policies aimed at reducing inequality among the regions.

The Lorenz curve is the most familiar graphical tool for examining the distribution of welfare. It is a plot of the cumulative fraction of the population starting from the poorest on the X-axis against the cumulative fraction of resources on the Y-axis. If resources were equally distributed, with everyone receiving the same, the Lorenz curve would be the 45 degree line usually labeled the line of perfect equality. Whereas in the case of complete inequality, with the richest person having everything, a Lorenz curve running along the X-axis with a right angle at (100,0) to terminate at (100,100) would be generated. Figures 1 and 2 below show the Lorenz curves for Uganda, using both adjusted and unadjusted figures. The Lorenz curves depicting the regional distribution are not presented here but were also examined.

For Uganda, the Lorenz curves depict high inequality of welfare among individuals. There is a great divergence from the line of perfect equality. Individuals in the lower 60 percent category of the population spend less than 20 percent of the total expenditure whereas the top 10% receive more than 50 percent of the total expenditure. The lowest quintile receives a very insignificant share, as shown by figures 1 and 2. From previous results, the ratio of the shares of the highest to the lowest quintile is about 27. In other words, the richest 20 percent of the population spend 27 times more than the poorest 20 percent of the population. A breakdown by region shows that Kampala has the least deviation from the line of perfect equality, relative to the rest. The reason for this has already been mentioned earlier. It has the largest concentration of workers and the most educated groups in Uganda. It also has the poorest groups in Uganda.

Fig. 1. The Lorenz curve using adjusted data.

Percentage of income

Percentage of receivers

The divergence is less compared to the other regions that are composed of a mixture of peasant farmers and civil servants. Table 8 below provides the decomposition of inequality in Uganda when divided into the five regions. All the four measures are presented. Based on the literature, these inequality measures may not always give the same result or rank of the regions but we can see that they display some general agreement in several ways.

Fig. 2. The Lorenz curve for Uganda using unadjusted data

Percentage of income

Percentage of receivers

The most common definition of the Gini-coefficient is in terms of the Lorenz diagrams, that is, the ratio of the area between the Lorenz curve and the line of perfect equality, to the triangle between this line. For Uganda as a whole, the Gini coefficient is 0.59 for the adjusted per capita expenditure and 0.58 for the unadjusted figures. This shows that welfare is unequally distributed among the population. These results are closely related to the findings by Opio (1997) which show an overall estimate of 0.594 using 1994 data. A clear indication therefore is that there has been no improvement in the distribution of welfare between these two periods. One can conclude that the benefits of growth in Uganda continue to be enjoyed by a few and the distributional aspects are insignificant. These findings are useful for policy on the grounds that an increase in growth without a better distribution of the benefits of growth will not result in improvement in welfare for the overall population. The extent of inequality seems lower at the regional level. The Gini coefficient for Kampala is 0.413 and 0.401, respectively, using adjusted and unadjusted figures, and it is smaller than for the other regions. The regional results found here are fully consistent with Opio (1997) and the reasons why Kampala still has a relatively better distribution hold. Interestingly, one still finds that the differences in per capita mean expenditure do not account for a substantial amount of overall inequality.

Table 8. Inequality measures, by region, Uganda 1997

		Northern	Western	Central	Kampala	Eastern	Uganda
Gini	Adjusted	0.544	0.512	0.426	0.413	0.567	0.596
	Unadjusted	0.532	0.501	0.416	0.401	0.554	0.589
Atkinson eps=0.5	Adjusted	0.153	0.141	0.185	0.237	0.158	0.212
	Unadjusted	0.143	0.127	0.204	0.207	0.148	0.204
eps=1	Adjusted	0.283	0.253	0.34	0.395	0.297	0.364
	Unadjusted	0.267	0.234	0.349	0.357	0.281	0.351
eps=2	Adjusted	0.465	0.543	0.513	0.685	0.462	0.691
	Unadjusted	0.437	0.472	0.655	0.616	0.429	0.672
S.d of logarithms	Adjusted	0.817	0.741	0.926	0.929	0.859	0.912
	Unadjusted	0.794	0.724	0.876	0.891	0.834	0.887
Coeff. of variation	Adjusted	0.935	1.092	1.028	1.483	0.929	1.498
	Unadjusted	0.883	0.946	1.382	1.274	0.869	1.431

SOURCE: Authors' computation from the data.

The other measure commonly used to describe inequality and welfare is the Atkinson coefficient. This index is based on a social welfare evaluation of income distribution. We shall not go through the rigorous derivation of this index here but briefly describe it (see Deaton, 1997 for derivation). Given a social welfare function, the Atkinson index is constructed by computing the equally distributed equivalent income of distribution. This is defined as the level of income per head which, if equally distributed, would give the same level of welfare as the existing distribution. It is expressed in the form of an additive social welfare function, with an inequality aversion parameter epsilon. The inequality aversion is always greater than zero and it measures the degree to which social welfare trade-offs mean living standards on the one hand for equality of the distribution on the other. The choice of a particular value for epsilon is a value judgement. As epsilon rises, more weight is attached to transfers at the lower end of the distribution and less weight to transfers at the top. For instance, if epsilon=2, and individual i is twice as well off as j, then the marginal social utility of additional x to j is one fourth the total marginal social utility of additional x to j. As epsilon tends to infinity, the marginal social utility of the poorest dominates over all the other marginal utilities, and policy is usually concerned with only the poorest. The results from the Atkinson index are presented in table 8 above. The indices for the regional distribution are equally presented in the table and we can see that they greatly reinforce earlier observations about the distribution of welfare between the regions.

It is generally seen that all the indicators of inequality used in this study point towards the same conclusion, namely, that there is great welfare inequality in Uganda. This is not surprising since earlier studies had concluded the same, although what remains a major policy issue is that there seems to be no improvement in welfare distribution over the period 1994 to 1997.

5. A Poverty Profile for Uganda

In this section another method of describing the distribution of welfare in Uganda is explored. This approach involves the use of a poverty line. A poverty line is simply a cut-off point below which somebody is said to be poor. It may take the form of expenditure, income, food or even basic needs. In this case, the standard definition of the poverty line adopted is a level of expenditure which is assumed to be the minimum amount required to attain a " decent" standard of living. All those people whose expenditures fall below this line are classified as poor. The use of poverty lines is justified on the grounds that the examination of the conditions faced by the poorest people is of particular importance to policy makers and others concerned with social welfare. Once the poor have been classified, we can ascertain their characteristics and these characteristics are very important for choosing policies aimed at targeting and reducing poverty.

Usually the choice of a poverty line is done arbitrarily but in this case, it is convenient to choose a line that classifies a certain percentage of the population as poor. In this case poverty lines will be used in order to examine the poorest 10 percent and 30 percent of the population. The poorest 10 percent can be considered as the core poor while the poorest 30 percent as just poor. In terms of per capita expenditure, the core poor spend less than 14321.1 Uganda shillings per month using adjusted figures and 9888 shillings using the unadjusted figures. The poor on the other hand spend less than 27265.3 Uganda shillings per month (adjusted) and 18193.19 shillings (unadjusted). In terms of expenditure, we also see that the core poor spend about 7 times less than the average expenditure by an average household.

The geographical distribution of poverty emerging from the results of the study is presented in table 9 below. The greatest percentage of the poor is found in the Eastern Region, followed by the Northern and Western regions. The geographical ranking of poverty is reasonably consistent with the rankings from other studies. In all other studies such as World Bank (1993, 1995), the Northern and Eastern regions have the highest percentages of the poor. However, one major change in ranking is noted. The Eastern Region is most subject to poverty, yet earlier findings showed that the North was probably the poorest region in Uganda. The reason for the change in ranking as noted earlier could be that the Eastern Region has been subject to the most chronic food insecurity over the past few years, leading to famine in some parts. This was mainly due to the drought that hit the area making the people more vulnerable. On the other hand, Central and Kampala regions which show the least levels of poverty are more food secure. The Northern Region has been consistently among the poorest regions because of the insecurity in the area which has distorted the production systems in the region. While there are large numbers of the poor in the Eastern Region, we cannot conclude, for example, that poverty there is any worse than it is in the Northern Region.

Prior findings in this study showed that female-headed households were not worse off in terms of welfare compared to the male-headed households. This is also true when we categorise households according to poverty groups. In fact male-headed households constitute about 67% and 88 % of the poorest category in the urban and rural areas, respectively. This is consistent with the previous findings, although no breakdown of poverty by marital status and age of the household head is done here. The general conclusion is that male-headed households are more likely to be poorer than female-headed households.

For purposes of policy analysis, it is also important that we examine the incidence of poverty among different types of workers. Information on the relationship between activity status of the household head and poverty is presented in table 9 below. According to table 9 there is a clear relationship between the activity status of the head and poverty. People from households with a head who is primarily employed in agriculture (farming) are more likely to be poor or very poor compared to those from households with heads working in the other sectors. In rural areas, agricultural employment is almost always available for people who lack other opportunities. In Uganda it is true that the majority of rural households are headed by someone whose primary occupation is agriculture. We also find that as the education of the household head increases, there is progressively lower dependency on agriculture. This is an expected finding and portrays the notion that simply raising wages of government workers will not help reduce poverty. Rather what is required is a policy aimed at improving agricultural production and aimed at the self-employed workers. In other words the private sector should be strongly supported. Raising agricultural incomes in Uganda is very necessary in reducing poverty. This policy recommendation is perhaps consistent with earlier recommendations included in the Poverty Eradication Action Plan published by the Government of Uganda (1997). The focus should now be on how exactly to raise these incomes. This can be an interesting subject for future research.

Table 9. General distribution of the poor

Sex	Poorest 10%	Poorest 30%	All Uganda
Female	16.3	16.6	19.8
Male	83.7	83.4	80.2
Distribution of the poor by regions
Region	Poorest 10%	Poorest 30%	All Uganda
Northern	24.5	25.1	20.1
Western	24.5	23.4	29.9
Central	9.2	15.9	19.5
Kampala	0.0	2.7	10.2
Eastern	41.8	32.9	20.2
Distribution of poor by occupation
Occupation	Poorest 10%	Poorest 30%	All Uganda
Farming	53.1	46.4	26.2
Comm. trade	4.1	5.4	16.7
Civil servant	7.1	10.5	25.6
Casual labourer	19.4	22.0	11.4
Others	16.3	15.6	20.1
Distribution of the poor by educational level
Education level	Poorest 10%	Poorest 30%	All Uganda
None	15.3	13.6	7.5
Never completed P/S	36.7	33.6	21.2
Completed primary	24.5	20.0	15.2
Post-primary	19.4	20.0	21.4
Post-secondary	2.0	2.0	3.6
Tertiary level	1.0	6.8	18.9
University or higher educ.	1.0	4.1	11.4

SOURCE. Authors' own computation from the data

Turning to education, apparently, substantial returns on education exist in Uganda. To shed more light on this, members of households whose head has had little formal education or none are more likely to be poor or very poor than those in households where the heads have some education. The educational distribution of the poor shows that a majority of the core poor either had no education at all or dropped out in early stages (never completed primary). Again we see that post-secondary education significantly reduces poverty in both rural and urban areas. The effect is higher in urban areas and in rural areas primary education is not associated with a significant decline in poverty. The returns on education found in this analysis are consistent with other studies, that is, there are substantial private returns to higher education but less returns for primary education.

To shed more light on the nature of these returns, education of the household head along with the primary employment of the head show that, even for heads whose primary employment is farming, poverty rates decline as education increases. A household whose head is employed in farming and who has post-secondary education is likely to be less poor than one whose head has no education but is working as a commercial worker. Increased education therefore causes a steeper decline in the rate of poverty, and there is a clear pay-off to education for agricultural households.

These problems of low education were compounded by the past reluctance of governments to build human capital through well-designed investments in education and show that there are significant deficiencies in the education provided. Lack of facilities may be one reason, but a more realistic one is that access to education is costly to poor households. It follows that if a head of household had no education, then it is most likely his children will not get access to it either. The private costs, however, have been noted to vary inversely with the demand for education. Turning to the household size in terms of the different poverty groups from table 10 we can see that poorer households are larger in size than wealthier ones in both rural and urban areas. This is both before and after adjusting for adult equivalence scales. The average household size for the core poor (poorest 10%) in the rural areas is 10.28 members while for the urban areas it is 9.09. For the top decile it is 4.4 members. The size of the household increases significantly along with the household poverty status, that is, the more you move to the poorer deciles the larger the household size. Overall, we see an increasing prevalence of poverty as the household size increases for all Uganda. This is consistent with previous studies by the World Bank (1995) and Opio (1997) that the likelihood of being poor increases as the number of people in the household increases. The rates of ownership and use of different types of household features (such as drinking water, lighting and cooking) for the poor and non-poor families are shown in table 10 below. The rural poor are likely to receive their drinking water from an unprotected well while the urban poor rely more on bore-holes. Poor households are also more likely to use firewood for cooking and less likely to use electricity for lighting. The rural poor use no electricity at all. Both rural and urban poor households depend on firewood for cooking. Observations and previous data suggest that fuelwood collection for use in households may put stress on Ugandan forests. Charcoal production for urban consumers is also strongly associated with increased deforestation. More research findings in this area are expected in the near future.

Table 10. Household characteristics by rural/urban distribution

Characteristics	Poorest 10%		Poorest 30%
Sex of head	Urban	Rural	Urban	Rural
Male	66.7	88.3	88.1	73.4
Female	33.3	11.7	11.9	26.6
Occupation
Farming	38.1	57.1	37.0	50.0
Comm. trade	14.3	1.3	12.4	2.8
Civil servant	19.1	3.9	18.5	7.5
Casual labourer	4.8	23.4	16.1	24.3
Others	23.8	14.3	16.1	15.4
Education
None	19.1	14.3	11.1	14.5
Never completed primary	38.1	36.4	33.3	33.6
Completed primary	19.