Republic of Namibia
CENTRAL BUREAU OF STATISTICS
NATIONAL PLANNING COMMISION
A REVIEW OF POVERTY AND INEQUALITY IN NAMIBIA
A Review of Poverty and Inequality in Namibia
October 2008
Central Bureau of Statistics
National Planning Commission
Executive summary
This report presents an analysis of poverty and inequality in Namibia based on the ex-
penditure data from the 2003/2004 Namibia Household Income and Expenditure Survey
(NHIES) conducted by the Central Bureau of Statistics. The main analytical purpose of
the report is to establish a new set of poverty lines for Namibia based on the Cost of Ba-
sic Needs (CBN) approach, which has become part of the poverty monitoring standard
among the Southern African Development Community (SADC) and most developing
countries. Poverty lines are particularly useful for drawing of poverty profiles, examining
the determinants of poverty and guiding policy interventions aimed at poverty reduction.
The process of setting the new poverty line can be split into two major stages. First, using
the NHIES data for households with low consumption expenditure, a food basket is de-
termined based on actual consumption patterns of low income households. Second, tak-
ing into account non-food requirements in addition to food needs, two poverty lines are
established for poor and severely poor households where consumption levels per
adult equivalent are lower than N$ 262.45 and N$ 184.56, respectively. Then by using
these definitions the incidence of poor and severely poor households are computed at
27.6 percent and 13.8 percent, respectively. Compared to the poverty line used previously
by the Central Bureau of Statistics, which was based on a simple relationship between the
food expenditure and total expenditure, the impact of the revised methodology for setting
the poverty lines is practically unchanged for poor households (previously estimated at
27.8 percent). However, the incidence of severely poor households is almost three times
higher under the new poverty line (previously estimated at 3.9 percent). Sensitivity tests
show how the new poverty lines are quite robust to small changes in specification while
the analysis provides further evidence as to the classification of high expenditure house-
holds as poor or severely poor under the food-share method.
Using the new CBN-based poverty lines, the study presents a detailed poverty profile of
Namibia. This profile shows that poverty status in the country is closely correlated with a
series of social, demographic, geographic and economic features of households. Multi-
variate analysis confirms that poverty levels in Namibia are higher for instance among
households that are female-headed, based in rural areas and have one or more children.
These results underscore the potential for poverty reduction through targeting of policies
and interventions. Consumption expenditure is positively correlated with the education
levels of the head of household. The higher the level of education, the higher the levels of
consumption expenditure and the more likely the household is to be classified as non-
poor holding other factors constant. These results underscore the centrality of strengthen-
ing the education system, especially expansion in access to secondary education, as an
important part of the national poverty reduction strategy. Relying on pension as a main
source of income is associated with lower levels of consumption expenditure and a higher
probability of poverty compared to other income sources. One way of explaining that
pensions are inadequate to lift households above the poverty line is that households that
rely on pensions as their main source of income are generally larger than other house-
holds. In other words, a greater number of people depend on the pension for their liveli-
hood than merely the pensioner. This type of information is important to consider when
determining the appropriate levels of social transfers and assessing their impact on pov-
erty. Differences in poverty levels also prevail according to administrative regions: the
Kavango and Ohangwena regions not only have the highest levels of incidence of poverty
but they are also home to the largest shares of poor households. These findings suggest
the potential for greater geographical targeting of anti-poverty programmes and for ensur-
ing that the benefits of the economic growth process accrue more favourably to these and
other disproportionately poor regions, when relevant policy interventions may be consid-
ered.
The analysis further reveals how unequal the consumption expenditure patterns are in
Namibia. The 10 percent of households with the lowest levels of expenditure account for
just over 1 percent of total expenditure in Namibia. The 10 percent of households with
the highest expenditure account for more than 50 percent of total expenditure. Stated in
another way the wealthiest 10 percent in the country have consumption levels that are 50
times higher than the poorest 10 percent. The Gini coefficient, which is the standard
summary measure for inequality, is 0.63 and with great variation according to various
background variables such as sex, age, main source of income and administrative region.
A comparison with countries for which comparable data is available suggests that the
level of inequality in Namibia is among the highest in the world. Additional measures of
inequality are introduced in order to deepen the understanding of inequality in Namibia.
Notably, a generalised entropy index is used for a decomposition exercise that reveals
how in general inequality in Namibia is the product more of inequality within different
social groups rather than of inequality between them. Nevertheless, between-group ine-
quality is sizeable especially when the population is arranged by main language spoken
and educational attainment. Moreover, the analysis introduces two measures of polarisa-
tion, which quantify the extent of the concentration of expenditure among distinct groups.
These results suggest that in addition to being among the most unequal societies in the
world, Namibia is also among the most polarised.
The report highlights a range of methodological aspects in the establishment of the pov-
erty line for Namibia and documents the technical steps involved. However, while the
process has been pursued with the greatest possible methodological rigour, eventually the
setting of any poverty line necessarily involves an element of subjectivity as to where ex-
actly the cut-off points in the distribution are put. Moreover, poverty is a dynamic phe-
nomenon of multiple dimensions, which goes beyond money-metric measures such as
income and consumption, which has been the main focus of this report. The analysis pre-
sented in this report must therefore not be regarded in isolation but as part of a broader
effort that relies on quantitative as well as qualitative approaches to contribute to the un-
derstanding of poverty in Namibia as an important basis for designing effective interven-
tions to improve the welfare of Namibians. Additional analysis also needs to be carried
out on the NHIES data to facilitate comparability with an earlier survey, study trends
over time in poverty and inequality, finalise the analysis of the income and nutrition data,
as well as more indepth analysis of regional aspects of poverty prevalence.
Foreword
This report contains one of the most comprehensive and authoritative analyses of poverty
and inequality conducted to date in Namibia. It is based on the 2003/2004 Namibia
Household Income and Expenditure Survey (NHIES) and includes the establishment of a
new poverty line based on the cost of basic food and non-food needs. This new method
replaces the approach used since the first NHIES in 1993/1994 and accordingly this re-
port presents new estimates of household poverty and an updated and expanded poverty
profile. Moreover, new measures for inequality and polarisation are introduced.
The work contained in this report is part of the Central Bureau of Statistics mission to
make available timely and high quality data based on definitions firmly anchored in in-
ternational best practices. The analysis was carried out by an in-house team of statisti-
cians with technical assistance in analysis and report-writing from Sebastian Levine
(UNDP, Namibia) and Benjamin Roberts (Human Sciences Research Council, South Af-
rica). The methodology and results presented in the report have been subject to extensive
scrutiny by stakeholders in Namibia and from international experts. Special acknowl-
edgement goes to Julian May (University of KwaZulu-Natal, South Africa) and Haroon
Bhorat (University of Cape Town, South Africa) who peer-reviewed the report and who
made many valuable observations that have significantly strengthened the final analysis.
The team is also grateful for many insightful comments from Jean-Yves Duclos (Univer-
sity of Laval, Canada), Erik Thorbecke (Cornell University, USA) and Abdelkrim Araar
(University of Laval, Canada).
The NHIES on which this report is based was financed with support from the Swedish
Agency for International Development, UNDP and UNICEF. Funding for the preparation
of the present report was made available by the Government of Namibia and the UNDP
Thematic Trust Fund for Poverty Reduction. Last but not least, we are sincerely grateful
for the support of everyone who helped make this report possible especially all the
households who participated in the 2003/2004 NHIES and indeed to the CBS dedicated
NHIES management and analysis team without whose inputs this report would not have
been possible.
FSM Hangula
Government Statistician
List of acronyms and abbreviations
AIDS Acquired Immune Deficiency Syndrome
Ave. Average
CBN Cost of Basic Needs
CBS Central Bureau of Statistics
CPI Consumer Price Index
DAD Distributive Analysis/Analysis Distributive
DER Duclos-Esteban-Ray
DRB Daily Record Book
Exp. Expenditure
FAO Food and Agricultural Organisation of the United Nations
FEI Food Energy Intake
FGT Foster-Greer-Thorbecke
GRN Government of the Republic of Namibia
HH/hh Households
HPI Human Poverty Index
NHIES Namibia Household Income and Expenditure Survey
No. Number
N$ Namibian Dollar (1 USD = 6.4 N$ in 2004)
NPC National Planning Commission
NDP National Development Plan
OLS Ordinary Least Squares
PES Post Enumeration Survey
PPP Purchasing Power Parity
PSU Primary Sampling Unit
RDA Recommended Dietary Allowance
SADC Southern African Development Community
SSD Social Sciences Division
Std. Dev. Standard Deviation
UNAM University of Namibia
UNDP United Nations Development Programme
UNSD United Nations Statistics Division
UNU United Nations University
US$ United States Dollar
VIP Ventilated Improved Pit
WHO World Health Organisation
Table of contents
1. Introduction................................................................................................................ 1
2. A new poverty line..................................................................................................... 2
3. Poverty profile ........................................................................................................... 6
4. Household assets and living conditions ................................................................... 18
5. Determinants of consumption and poverty .............................................................. 29
6. Inequality and polarisation....................................................................................... 33
7. Conclusion ............................................................................................................... 39
References......................................................................................................................... 40
ANNEXES........................................................................................................................ 42
ANNEX A: Background to the NHIES data
ANNEX B: Poverty measures in Namibia and SADC
ANNEX C: Setting the poverty line
ANNEX D: The national food basket
ANNEX E: First-order stochastic dominance tests
ANNEX F: Poverty profile tables
ANNEX G: Multivariate analysis
ANNEX H: Measures of inequality and polarisation
ANNEX I: Confidence intervals
- 1 -
1. Introduction
The reduction of poverty and inequality remains an overarching priority for the Govern-
ment of Namibia. The national development framework, Vision 2030, finds that in the
current situation: Inequality and poverty endangers social harmony, peace and democ-
racy and sets as its long-term development objective: Poverty is reduced to the mini-
mum, the existing pattern of income-distribution is equitable and disparity is at the mini-
mum. (GRN, 2004: 104f). Moreover, strategies to implement Vision 2030, such as the
successive medium-term National Development Plans, the 1998 Poverty Reduction Strat-
egy and the 2001 National Poverty Reduction Action Programme all have reduction of
poverty and inequality among their chief objectives (GRN 1998, 2001, 2002, 2005). Na-
mibia is also signatory to major agreements that shape the global development policy
agenda; notably the 2000 Millennium Declaration, which commits countries to cut the
1990 incidence of income poverty by half before 2015 and a range of other social devel-
opment objectives known as the Millennium Development Goals.
This report presents and applies a new approach to defining poverty levels by presenting
a new set of poverty lines, which is rooted in an absolute measure of wellbeing linked to
a minimum required nutritional intake. Such a Cost of Basic Needs (CBN) approach is
becoming standard among statistical agencies throughout SADC and other developing
countries but it has not been used for official statistics in Namibia before. A key feature
of the analysis presented in this report is that it is based on international best-practices
combined with an extensive national consultative process that has created broad consen-
sus and ownership of the approach. It should be noted that the present report focuses on
poverty as measured through the extensive expenditure data from the 2003/2004 Namibia
Household Income and Expenditure Survey (NHIES) and therefore relies heavily on a
money-metric approach to poverty measurement. Since poverty more generally is under-
stood to be a complex phenomenon of multiple dimensions that go beyond the lack of
income and money, the present analysis should be considered complementary to other
approaches that focus on poverty in other domains. Moreover, the quantitative methodol-
ogy outlined here could be fruitfully combined with qualitative approaches for a fuller
understanding of poverty, its determinants and ways to overcome it.
The report is organised as follows: After this Introduction, Section 2 gives a short intro-
duction to the new poverty line for Namibia; Section 3 presents a poverty profile with
details on levels of poverty according to a range of economic, social and demographic
variables and in Section 4, results from an analysis of household access to various facili-
ties and ownership of assets are presented. In Section 5, some key determinants and driv-
ers of poverty are discussed and in Section 6 issues of inequality and polarisation in the
distribution of household expenditure are reviewed. Finally, Section 7 concludes. The
intention has been to keep the main text as non-technical as possible in order to make it
user-friendly for a wide audience. However, since a poverty assessment of this nature in-
variably involves a series of technical elements and methodological decisions, which
need to be taken into account when discussing the results, the report includes a series of
annexes with more detailed documentation of the approaches and results.
- 2 -
2. A new poverty line
The poverty line used in this report differs from previous practices in Namibia. In the
past, the official poverty line was defined using the relative share of food expenditure to
total expenditure of households.1 This way a household was considered poor if food
expenditure made up 60 percent or more of total expenditure. The household was classi-
fied as severely poor if food expenditure made up 80 percent or more of total expendi-
ture. While it is generally accepted that the share of food expenditure rises with falling
total expenditure, there are a number of methodological problems with the approach, es-
pecially when it comes to identifying the poorest households, and determining the cut-off
points in the welfare distribution (see Annex B). In place of the food-share method, a
more direct method to setting a poverty line is therefore adopted. This methodology is
often referred to as the Cost of Basic Needs (CBN) approach and is used widely in the
SADC region and in developing countries more generally. Under this approach the pov-
erty line is set by first computing the cost of a food basket enabling households to meet a
minimum nutritional requirement and then an allowance for the consumption of basic
non-food items is added. Households with consumption expenditure in excess of this
threshold are considered non-poor and households with expenditure less than the thresh-
old are considered poor. The principal reason that the Central Bureau of Statistics uses
consumption expenditure instead of income data is that household earnings can be highly
irregular over time while expenditures tend to be more stable. Moreover, income is likely
to be underreported for some groups and consumption measures are able to better capture
the contribution from informal activities and own production, which make significant
contributions to household welfare especially in developing countries and certainly in
Namibia.
Setting up an absolute poverty line for Namibia using the CBN approach has been a fairly
labour intensive process and has included a series of methodological steps. These steps
are detailed in Annex C. In summary, the process of setting the poverty line began by es-
tablishing a food basket, which was determined by the actual consumption patterns of the
households with low consumption levels (Annex D provides details of the contents of the
food basket for purchased and in-kind items). The monetary value of attaining a mini-
mum nutritional intake of 2,100 kcal in a low-income household was then computed
based on available prices taking into account regional price differences, and this value
then formed the food poverty line (N$ 127.15). While having sufficient resources in the
household to meet food requirements is critical, it is not enough to classify a household as
poor or non-poor. This is so because households that can afford to meet the food re-
quirements of all members but lack resources to purchase clothing and shelter, for exam-
ple, should be considered deprived in a very basic sense.
Two approaches for estimating the non-food components of the poverty line were used in
the analysis: 1) Under the first approach, non-food expenditure was calculated from ac-
tual expenditure on non-food items by households where food expenditure is approxi-
1 See details of the NHIES survey instrument in Annex A. See Annex B for an extensive overview of past
and present poverty measures used in Namibia and other SADC countries.
- 3 -
mately equal to the food poverty line. This component is then added to the food poverty
line. 2) Under the second approach, non-food expenditure is calculated from actual non-
food expenditure of households whose total expenditure is equal to the food poverty line.
Similarly, this component is then added to the food poverty line. The rationale for the lat-
ter, more austere approach is that if these households are able to obtain minimum food
basket, but choose to divert resources to buy non-food items, then the household must
clearly view these items as essential.
In the literature on poverty measurement, both methods are found to be methodologically
sound and they are often considered together as a lower and upper bound, respectively
(Ravallion 1998). In the subsequent poverty analysis for Namibia both measures are ap-
plied and should be interpreted as representing a range in poverty levels in the country.
Households that have consumption expenditures below the upper bound poverty line are
classified as poor and those with consumption expenditures below the lower bound
poverty line are classified as severely poor. Those households with consumption ex-
penditure above the upper bound poverty line are considered non-poor. Table 1 shows
the values of the food poverty line as well as the upper and lower poverty lines for the
2003/2004 NHIES and Figure 1 illustrates the upper and lower poverty line in the actual
distribution of household expenditure. The figure also illustrates how in the definition of
poor and severely poor, the latter is a subset of the former. The values of the poverty
lines are expressed for households but in adult equivalents thus adjusting for differ-
ences in the age composition of household members (see Annex C for more details on
this adjustment).
Table 1: Annual values of national poverty lines, monthly N$ per capita
Poverty line 2003/2004
Food poverty line 127.15
Lower bound poverty line: severely poor 184.56
Upper bound poverty line: poor 262.45
Once the poverty lines have been determined, the final step is to select the measures to
express the shortfall and deprivation. As has become standard in poverty research, the
analysis presented for Namibia follows Foster, Greer and Thorbecke (1984) by using the
most common of the so-called Foster-Greer-Thorbecke (FGT) class of poverty measures.
These are:
" The headcount index or incidence of poverty. This is the most commonly used
and the easiest of the three measures to interpret. It shows the proportion of the
population or households that are below a given poverty line and is usually ex-
pressed as a percentage of the total population or number of households.
" The poverty gap index measures the depth of poverty given by the distance or gap
between actual total expenditure of poor households and the poverty line. This
- 4 -
measure can be thought of as the percentage of the poverty line needed to bring
those below the threshold up to the poverty line.
" The poverty severity index gives more weight to the shortfall in consumption ex-
penditure further below the poverty line. This index is thus sensitive to the ine-
quality among the poor. The index will rise with inequality within the group of
poor.2
Household expenditure
That poorer households divert a substantial share of their total expenditures to food is
evident from Table 2, which presents a breakdown of expenditures for different catego-
ries of expenditure by consumption expenditure quintiles. The table shows how the rela-
tive share of food expenditure falls as expenditure increases. Among the 20 percent of
households with the lowest consumption expenditure (quintile I), 56.7 percent of total
expenditure is devoted to food compared to just 13.2 percent in the 20 percent of house-
holds with the highest consumption expenditure (quintile V).
The share of expenditure devoted to housing and utilities, and clothing and footwear, is
rather stable across the distribution. For remaining expenditure categories, the shares in-
crease with expenditure. For instance, the share of expenditure devoted to transportation
in the 20 percent of households with the lowest consumption expenditure is 2.3 percent
compared to 19.9 percent in the 20 percent of households with the highest consumption
expenditure. This pattern can be explained by the larger share of subsistence farmers and
pensioners among the poorest households (explored further below) while non-poor
households are often wage earners who are more likely to have commuting needs and
certainly have a greater degree of ownership of motor vehicles. Expenditure shares on
education and health are also double or more in the wealthier households. This is proba-
bly the result of a combination of factors including the waiver system for the payment of
School Development Fees for the poorest households and the ability of wealthier house-
holds to afford private education. For the poorest 20 percent of households, a total of 80
percent of expenditure is devoted to food and shelter. The third largest share is devoted to
clothing and footwear and thus only minor shares are available for education and health
care.
In summary, while the expenditure patterns of the wealthier households are more bal-
anced across the expenditure categories, the groups with the lowest levels of consumption
expenditure concentrate most of that expenditure covering basic needs especially food.
However, these households also divert a significant share of their expenditure towards
non-food items. As described above, the CBN approach to setting the poverty line en-
sures that both food and non-food items are catered for when determining the basic needs
of households.
2 See Annex C for more details on the FGT measures.
- 5 -
Figure 1: Distribution of NHIES expenditure and the poverty lines, 2003/2004
Per adult equivalent monthly expenditure, N$
1,0009008007006005004003002001000
N
u
m
b
e
r
o
f
h
o
u
s
e
h
o
ld
s
40,000
30,000
20,000
10,000
0
Note: Horisontal axis truncated to the right to enhance clarity.
Table 2: Expenditure shares by quintiles
Quintiles of adult equivalent expenditure
Annual household expenditure on: I II III IV V Total
Food 56.7 54.8 46.8 33.5 13.2 26.3
Housing, including utilities 23.4 21.0 20.4 20.6 24.4 23.0
Transport 2.3 2.9 4.9 9.9 19.9 14.1
Furniture and equipment 3.7 4.9 7.4 8.9 10.6 9.1
Clothing and footwear 6.6 7.6 8.0 8.7 5.1 6.3
Recreation, entertainment and sport 0.5 0.8 1.2 2.1 5.0 3.5
Communication 0.8 1.3 2.0 3.0 3.9 3.1
Education 1.5 1.3 1.8 2.5 3.6 2.9
Health care 1.2 1.3 1.4 1.7 2.2 1.9
Accommodation services 0.1 0.1 0.2 0.2 0.7 0.5
Miscellaneous expenditure 3.1 4.0 5.7 8.9 11.3 9.2
Total 100.0 100.0 100.0 100.0 100.0 100.0
Number of households in sample 1,904 1,889 2,009 2,143 1,856 9,801
Weighted number of households 74,306 74,376 74,344 74,304 74,346 371,678
Lower bound poverty line, N$ 185.56
Upper bound poverty line, N$ 262.45
- 6 -
3. Poverty profile
In this section, the poverty lines for those that are poor and those that are severely
poor are used to draw a consumption expenditure-based income poverty profile for Na-
mibia. This profile describes the two overlapping categories of poor households accord-
ing to a range of economic, social and demographic variables, and makes comparisons
with the category of non-poor households. In the bi-variate analysis, the poverty status
of households is compared with background variables one by one. This type of analysis is
particularly suited for identifying where the poor live and is important for targeting of
poor households. A subsequent section will use multivariate analysis to account for the
simultaneous effects of several variables and explore the determinants of poverty. It
should be noted that the poverty profile is purely descriptive and that causality cannot be
inferred from the correlations. Simplified tables and graphs have been included to bring
out some of the main results but a more detailed set of tables are included in the Annex F.
Figure 2: Changes in poverty levels as a result of the revised poverty line, 2003/2004
0% 10% 20% 30%
New poverty line
(Cost of basic
needs)
Old poverty line
(Food share ratio)
Poor
Severely poor
Poor 27.6% 27.8%
Severely poor 13.8% 3.9%
New poverty line (Cost of basic
needs)
Old poverty line (Food share ratio)
Note: Under the old poverty line poor households where identified as those spending 60 percent or more
of their total expenditure on food, and severely poor households as those spending 80 percent or more.
Under the new approach to setting a poverty line poor households are those that have monthly expendi-
tures of less than N$ 262.45 per adult equivalent, and severely poor households as those with expendi-
tures of less than N$ 184.56.
- 7 -
Figure 2 compares the poverty incidence that resulted from the old method of setting the
poverty line using the food-share ratios of 60 and 80 percent for poor and severely poor,
respectively, with the new measure based on the Cost of Basic Needs (CBN) approach.
As can be seen, the two methods arrive at very similar results when it comes to the inci-
dence of poor; 27.8 for the food-share ratio (60%) and 27.6 for the new CBN-based pov-
erty line (upper bound). However, there are important changes in the composition of the
poor households, e.g. the share of urban poor households almost doubles using the new
measure. Moreover, the revision in the method for setting the poverty line has a signifi-
cant impact on the classification of severely poor. Using the food-share method (80% and
above), 3.9 percent of households are classified as severely poor whereas the share is
more than three times higher, 13.8 percent, under the new method using the (lower
bound) poverty line. Clearly this does not imply that the incidence of severely poor has
increased over time but simply that in comparison, the old method underestimated the
incidence of the poorest among the poor. In effect, these new figures represent revisions
of the official poverty figures for Namibia. The figure also illustrates how the group of
severely poor form a sub-set of the poor. Therefore, in general when reference is made to
poor households these also include severely poor ones.
Table 3: Incidence, depth and severity of poverty by households, 2003/2004
Incidence (P0) Depth (P1) Severity (P2)
Poor 27.6 % 8.9 % 4.1
Severely poor 13.8 % 3.9 % 1.7
Table 3 presents the results of the three FGT measures using the two poverty lines for
poor and severely poor households. As noted already, the incidence of poor households is
27.6 percent and 13.8 percent for severely poor households. In these households the adult
equivalent expenditures are too low to cover the basic food and non food needs on which
the poverty lines are based. The depth of poverty among poor households is 8.9 percent,
which indicates that on average households are 8.9 percent below the upper bound pov-
erty line. Similarly, households are on average 3.9 below the lower bound poverty line.
The severity of poverty gives a higher weight to the poorest of the poor and this measure
is particularly useful in tracking developments over time and comparing deprivation be-
tween regions. In the detailed tables in Annex F, the three measures of poverty are pre-
sented for each of the poverty lines.
Poverty incidence by sex and age
Figure 3 shows the correlation between the incidence of poverty and the sex of the head
of household. Among households headed by females, 30.4 percent are poor and 15.1 per-
cent are severely poor. This is higher than for male-headed households where 25.8 and
12.9 percent are poor and severely poor, respectively. In Annex I, confidence intervals
using the conventional levels are reported for most of the poverty incidence estimates.
Given the overlapping values for male- and female-headed households when it comes to
the lower bound poverty line, it can be concluded on the basis of Table I-3 that there is no
significant difference in the incidence of severely poor households by sex of the head of
- 8 -
household. However, with the confidence intervals not overlapping when it comes to the
upper bound poverty line, it can be concluded that the incidence of poor households is
significantly higher among those headed by females. The multivariate analysis later con-
firms that when controlling for differences in education, sources of income and other fac-
tors female-headed households have lower incomes and face a higher probability of being
poor than male-headed households. Moreover, the gender differences may be even
greater in reality as the NHIES data do not reveal potentially important inequalities
within the household.
Figure 3: Incidence of poverty by sex of head of household, 2003/2004
10% 20% 30% 40%
Total
Female
Male
Poor
Severely poor
Poor 27.6% 30.4% 25.8%
Severely poor 13.8% 15.1% 12.9%
Total Female Male
Differences in poverty status and age levels of the head of household are presented in
Figure 4. Among those households with heads of household aged 16-20 years, 22.5 per-
cent are poor and 14.4 percent are severely poor. This compares with the national average
for all age groups of 27.6 percent poor and 13.8 severely poor. Those aged 30-34 have
the lowest shares of poor and severely poor, 17.9 and 7.5 percent, respectively. From then
on, the share of poor increases steadily. Among those aged 65 and older, the incidence of
poor is 47.5 percent and the incidence of severely poor is 22.7 percent. The average age
of heads of households in Namibia at the time of the survey was just under 47 years.
However, the average age of the household head among poor households was 53 years
compared to 44 years among non-poor households. One hypothesis that could explain the
differences in poverty levels by age groups is that those in the ages 25-39 are more likely
to hold salaried jobs, which in turn is associated with a lower probability of household
poverty (further explored below). Moreover, at higher age levels household heads are of-
ten reliant on a pensions as a main income source, which in turn is an important determi-
- 9 -
nant of higher probability of the household being poor (also explored further below). The
levels of depth and severity of poverty are also higher for the older age groups.
Figure 4: Incidence of poverty by age of head of household, 2003/2004
0% 10% 20% 30% 40% 50%
Total
16-20
21-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65+
Poor
Severely poor
Poor 27.6% 22.5% 19.1% 18.2% 17.9% 18.7% 23.1% 22.4% 27.0% 35.4% 42.6% 47.5%
Severely poor 13.8% 14.4% 10.8% 8.3% 7.5% 10.0% 12.4% 12.1% 12.1% 18.3% 23.6% 22.7%
Total 16-20 21-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65+
Poverty incidence by locality and region
Poverty incidence varies greatly between the administrative regions of Namibia and be-
tween urban and rural areas as reflected in Figure 5 and Figure 6. The incidence of poor
households in the rural areas is 38.2 percent compared to 12 percent in urban areas.
Moreover, 19.1 percent of households in rural areas are severely poor compared to 6 per-
cent in urban areas. As indicated by Table I-1 in Annex 1, these differences are statisti-
cally significant. Among the regions, the highest incidence of poverty is in the Kavango
region where 56.5 are poor and 36.7 percent are severely poor. In Ohangwena, the inci-
dence of poor and severely poor households is 44.7 and 19.3 percent, respectively. Pov-
erty incidence is lowest in Khomas and Erongo with 6.3 and 10.3 percent, respectively.
The measure for the depth of poverty is 23 percent in Kavango and 13.1 percent in Har-
dap (see Table F-1 in Annex F).
- 10 -
Figure 5: Incidence of poverty by locality of household, 2003/2004
0% 10% 20% 30% 40% 50%
Total
Urban
Rural
Poor
Severely poor
Poor 27.6% 12.0% 38.2%
Severely poor 13.8% 6.0% 19.1%
Total Urban Rural
Figure 6: Incidence of poverty by region, 2003/2004
0% 10% 20% 30% 40% 50% 60%
Total
Khom
Erong
Oshan
Karas
Kunen
Otjozo
Capri
Omah
Omus
Harda
Oshik
Ohang
Kavan
Poor
Severely poor
Poor 27.6% 6.3% 10.3% 19.6% 21.9% 23.0% 27.8% 28.6% 30.1% 31.0% 32.1% 40.8% 44.7% 56.5%
Severely poor 13.8% 2.4% 4.8% 7.8% 12.5% 13.1% 15.8% 12.5% 17.5% 12.8% 21.9% 16.6% 19.3% 36.7%
Total Khom Erong Oshan Karas Kunen Otjozo Capri Omah Omus Harda Oshik Ohang Kavan
- 11 -
Given the differences in the sizes of the populations between the regions, it is useful to
look at the poverty shares in addition to the incidence of poverty. The poverty share is
computed based on the total number of poor households and poverty shares by region are
presented on Figure 7. The figure shows that Kavango and Ohangwena regions not only
have the highest levels of incidence of poverty, but they also have the largest shares of
poor households. Those two regions are home to 17.8 and 16.5 percent, respectively of all
the poor households in Namibia. In other words, of all poor households in Namibia, more
than one third live in Kavango and Ohangwena. Add Oshikoto and Omusati and those
four regions combined account for almost 60 percent of all poor households in the coun-
try.
Additional tests on the sensitivity of the poverty lines show that the ranking of Kavango
and Ohangwena as the regions with the highest incidence of poverty is quite robust to
changes in the value of the poverty line. The same goes for Khomas and Erongo, which
are ranked lowest when it comes to poverty incidence. However, the ranking of other re-
gions is more sensitive to the specification of the poverty line (see Figure E-5 in Annex
E). This is important in such cases where planning decisions and budget allocations are
made on the basis of ranking of regions and for which special care should be taken in as-
certaining the robustness of such rankings.
Figure 7: Poverty shares by region, 2003/2004
Karas
3.3%
Khomas
4.0%
Erongo
2.8%
Oshana
6.1%
Kunene
3.0%
Otjozondjupa
7.8%
Caprivi
5.2%
Omaheke
3.9%
Omusati
11.9%Hardap
5.1%
Oshikoto
12.7%
Ohangwena
16.5%
Kavango
17.8%
Poor households tend to be larger in terms of the number of people than non-poor house-
holds. Severely poor households are even larger. Table 4 shows that the average house-
hold size is Namibia is 4.9 persons with 4.2 on average in urban areas and 5.4 in rural
areas. Among households classified as poor, the average household size is 6.7 compared
to 4.2 for non-poor households. For households that are severely poor, the average size is
- 12 -
7.2. Households are also bigger on average among rural poor than among urban poor.
Highest above the national average are severely poor households in Kunene with an aver-
age of 8.6 household members. The farthest below the national average are the non-poor
households in Erongo with an average of 3.4 household members.
Table 4: Average size of households by region and locality according to poverty
status, 2003/2004
Severely poor Poor Non-poor Total
Caprivi 6.4 5.9 4.1 4.6
Erongo 5.7 5.0 3.4 3.6
Hardap 5.8 5.5 3.6 4.2
Karas 6.9 6.0 3.5 4.0
Kavango 7.7 7.3 5.3 6.4
Khomas 5.7 5.2 3.9 4.0
Kunene 8.6 7.4 3.8 4.6
Ohangwena 8.4 7.8 5.0 6.3
Omaheke 6.5 5.8 3.5 4.2
Omusati 7.0 7.1 5.1 5.7
Oshana 7.2 7.0 5.0 5.4
Oshikoto 7.0 6.5 4.6 5.4
Otjozondjupa 6.8 6.1 3.7 4.3
Namibia 7.2 6.7 4.2 4.9
Urban 6.5 6.0 4.0 4.2
Rural 7.3 6.9 4.5 5.4
Table 5: Average number of children under 18 in households by region and locality
according to poverty status, 2003/2004
Severely poor Poor Non-poor Total
Caprivi 3.5 3.1 1.9 2.2
Erongo 2.9 2.2 1.1 1.2
Hardap 3.0 2.7 1.4 1.8
Karas 3.3 2.8 1.1 1.5
Kavango 4.2 4.1 2.6 3.5
Khomas 2.3 1.8 1.2 1.3
Kunene 5.1 4.1 1.7 2.2
Ohangwena 4.9 4.5 2.6 3.5
Omaheke 3.3 2.9 1.3 1.8
Omusati 3.8 3.9 2.7 3.0
Oshana 3.7 3.7 2.3 2.6
Oshikoto 3.8 3.6 2.3 2.8
Otjozondjupa 3.5 3.0 1.5 1.9
Namibia 3.9 3.6 1.8 2.2
Urban 3.0 2.6 1.4 1.6
Rural 4.1 3.8 2.2 2.8
- 13 -
Poorer households tend to be larger because they are home to more children than non-
poor households. Table 5 shows the average number of children under the age of 18 by
household poverty status in the various regions and localities of the country. Among all
households in Namibia, the average number of children is 2.2. Among non-poor house-
holds, the number is 1.8 and double, 3.6, in poor households. In households classified as
severely poor, the average number of children is even higher at 3.9. The lowest average
number of children per household is found among non-poor households in Erongo and
Karas where the average number is 1.1. The highest number is among severely poor
households in Kunene where there are an average of 5.1 children in the household.
Poverty incidence by language group
In the NHIES, respondents are asked about the main language spoken in the household.
Figure 8 presents the results of poverty incidence by language groups. Among those
households where Khoisan is the main language spoken, the incidence of poverty is 59.7
percent and the incidence of severe poverty is 39 percent or more than double the na-
tional averages. Similar high levels of both poverty and severe poverty are found among
speakers of Rukavango languages. Moreover, Khoisan and Rukavango-speaking house-
holds have the highest values for poverty depth (see Table F-3 in Annex F). Households
where the main language is Khoisan are on average 24.9 percent below the threshold for
poor households. For Rukavango-speaking households, it is 21.8 percent. Households
with Nama/Damara as the main language also have an incidence and depth of poverty
that is significantly above the national average. Conversely, the levels of poverty among
households where English and German are the main languages are less than 1 percent.
Figure 8: Incidence of poverty by main language spoken in household, 2003/2004
0% 10% 20% 30% 40% 50% 60% 70%
Total
Germ
Engl
Afri
Sets
Othe
Otji
Capr
Oshi
Nama/
Ruka
Khoi
Poor
Severely poor
Poor 27.6% 0.0% 0.6% 7.9% 14.5% 16.4% 17.0% 24.6% 28.5% 34.2% 54.4% 59.7%
Severely poor 13.8% 0.0% 0.4% 3.5% 1.0% 9.6% 8.8% 10.8% 11.8% 21.4% 34.9% 39.0%
Total Germ Engl Afri Sets Othe Otji Capr Oshi Nama/ Ruka Khoi
- 14 -
Another way of looking at the poverty levels among the language groups is by poverty
share, which takes into account the size of the population groups and indicates how much
each group contributes to the total number of poor. Poverty shares by language group are
presented in Figure 9. This way even if Khoisan has the highest incidence of poverty, it is
a relatively small group, less than 5000 households, and thus the group as a whole ac-
counts for 2.9 percent of all the poor households in Namibia. On the other hand, even if
poverty incidence in Oshiwambo-speaking households is 28.5 percent, and thus just
above the average for Namibia, since it is the largest of all the language groups, it also
has the highest share of all poor in the country, 50.5 percent. The fact that the language
groups differ tremendously in size as well as in their level of deprivation is important for
policy makers since reducing overall levels of poverty among the smaller more deprived
groups will require more targeted efforts compared to more broad-based initiatives to re-
duce poverty.
Figure 9: Poverty shares by main language spoken in household, 2003/2004
Otjiherero
5.4%
Caprivian
4.7%
Oshiwambo
50.6%
Nama/Damara
14.2%
Rukavango
18.4%
Khoisan
2.9%
Setswana
0.2%
Others
0.6%
German
0.0%
English
0.0%
Afrikaans
3.0%
Poverty incidence by education and income source
The results of the poverty profile provide further evidence to the critical role of education
in explaining poverty status of households. Figure 10 shows the incidence of poverty by
educational attainment of the head of household. Among those with no formal education,
50 percent are poor and 26.7 percent are severely poor. On average, these households
have total consumption expenditure levels that are 17.2 percent below the national
threshold for poor households. The situation improves as education levels increase.
Among those who have finalised their secondary education, 12.6 or less than half the na-
tional average, are poor and 5.1 percent are severely poor. Poverty among those who hold
a tertiary degree is virtually non-existent. Of all poor households, 83.5 percent have a
- 15 -
head of household that has either no formal education or has only completed primary
school.
Figure 10: Incidence of poverty by education attainment of head of household,
2003/2004
0% 10% 20% 30% 40% 50% 60%
Total
No formal
education
Primary education
Secondary
education
Tertiary education Poor
Severely poor
Poor 27.6% 50.0% 35.5% 12.6% 0.4%
Severely poor 13.8% 26.7% 17.7% 5.1% 0.1%
Total
No formal
education
Primary
education
Secondary
education
Tertiary
education
Since the poverty measure used in this analysis is based on a consumption-based measure
of poverty, it is closely associated with occupation and the main source of income for
households as shown in Figure 11. Households that rely on salaries and wages as their
main source of income have an incidence of both poor and severely poor that is less than
half of the national average. Still, since this group is so large46 percent of all house-
holds have salaries and wages as their main source of incomethat it makes up 23.1 per-
cent of all poor. In other words, a salaried income is by no means a guarantee of a life
above the poverty line in Namibia. Among households that rely on subsistence farming as
their main source of income, 40.3 percent are poor and 17.6 percent are severely poor.
These households also make up 42.3 percent of all poor households.
Among those relying on pensions as their main source of income, 49.6 percent are poor
and 28.4 percent are severely poor. These households are larger with an average of 5.3
people in them and thus a greater number of people other than the pension recipient rely
on the pension as the main source of income. Households where the main source of in-
come is salaries and wages have 4.2 members and the head of these households is on av-
erage 39.5 years. Unsurprisingly, households that rely on pensions are generally older
on average the head of these households is 69.3 years oldcompared to the national av-
erage of 46.9 years (Table 6).
- 16 -
Figure 11: Incidence of poverty by main source of income, 2003/2004 (%)
0% 10% 20% 30% 40% 50% 60%
Total
Salaries and
wages
Subsistence
farming
Household
business
Pensions
Poor
Severely poor
Poor 27.6% 13.8% 40.3% 24.1% 49.6%
Severely poor 13.8% 6.6% 17.6% 13.7% 28.4%
Total
Salaries and
wages
Subsistence
farming
Household
business
Pensions
Table 6: Household main income source and average age of households
head and average size
Main source of income
Average age of head of
household
Average house-
hold size
Pensions 69.3 5.3
Non-Farming Business Activities 40.4 4.7
Subsistence Farming 54.7 6.2
Salaries/Wages 39.5 4.2
All sources 46.9 4.9
Namibia is home to a growing number of orphans principally due to the increased mortal-
ity associated with the AIDS epidemic (Ministry of Health and Social Services 2001).
According to estimates based on the NHIES, a total of 85,000 households have either a
single or double parent orphan aged 0-17 years (i.e. one or both biological parent(s)
is/are not alive). Figure 12 shows the incidence of poor and severely poor by categories
of households with and without children and with and without orphans. The incidence of
poverty among households where there is at least one orphan is 41.8 percent, compared to
the national average of 27.6 and to 9.4 percent in households without any children aged
0-17. The share of severely poor households is 21.1 percent among households with at
least one orphan, compared to 13.8 percent for all households and 3.9 percent among
- 17 -
households without any children. Households with children also have higher levels of
poverty incidence even if these children are not orphaned. No conclusions about causa-
tion can be drawn on the basis of this analysis, e.g. that poverty leads households to have
more children or that households are poor because they have more children, but it can be
concluded that the presence of children and especially orphans should be highly effective
criteria in public policy interventions that aim to target poor households.
Figure 12: Incidence of poverty for households with children and orphans,
2003/2004 (%)
0% 10% 20% 30% 40% 50%
No children 0-18
years
No orphan 0-18 years
All households
Children 0-18 years,
not orphaned
Children 0-18 years
Orphan 0-18 years
Poor
Severely poor
Poor 9.4% 23.4% 27.6% 31.4% 34.6% 41.8%
Severely poor 3.9% 11.7% 13.8% 16.1% 17.7% 21.1%
No children 0-
18 years
No orphan 0-
18 years
All
households
Children 0-
18 years, not
orphaned
Children 0-
18 years
Orphan 0-18
years
- 18 -
4. Household assets and living conditions
The following analysis shows how monetary poverty is correlated to deprivation in a
range of other domains, including household assets, distance and access to facilities,
physical housing features and utilities, and other living conditions. The section under-
scores the multi-dimensional aspects of poverty in that deprivation in one dimension is
often associated with deprivation in other dimensions. This analysis also focuses on one-
to-one relationships between poverty status and different household conditions, and again
the focus is on simple statistical correlations and not underlying causes.
Household assets
Table 7 shows the correlation between the ownership of a range of household and agri-
cultural assets and the level of consumption expenditure in the household. Generally, the
higher level of consumption expenditure, the higher the share of households that own a
particular household asset. For instance, 60.4 percent of households in quintile I (i.e. the
20 percent of households with the lowest consumption expenditure) own a radio com-
pared to 86.1 percent in quintile V (i.e. the 20 percent of households with the highest
consumption expenditure). Moreover, 4.0 percent of households in the lowest quintile
own a refrigerator compared to 79.5 percent in the highest quintile. Similarly, only 1.5
percent of households in the lowest quintile own a motor vehicle compared to 60.6 per-
cent in the highest quintile. In quintile V, 25 percent of households own a computer while
for all other quintiles it is less than 2 percent.
Table 8 compares the ownership of and access to various agricultural assets across the
categories of severely poor, poor and non-poor households. Among the non-poor house-
holds, 34.2 percent own cattle and 37.6 percent own goats. Among the poor and severely
poor, 32.4 and 29.7 percent, respectively own cattle. Ownership of goats is higher among
poor and severely poor households than among non-poor households. Ownership of field
for crops is also higher among poor and severely poor households, 34.7 and 35.4 percent,
respectively. The communal land tenure system that is dominant in the northern regions
of Namibia explains the large proportions of both poor and non-poor households, 29.1
percent, that do not own but have access to land. Ownership of and access by households
to a plough is higher among poor households compared to both severely poor and non-
poor households.
- 19 -
Table 7: Asset ownership by quintiles of monthly expenditure per adult equivalent
Quintiles of adult equivalent expenditure
Owns asset I II III IV V Total
Household assets
Radio 60.4 67.6 68.9 74.1 86.1 71.4
Stereo HiFi 3.5 8.6 15.2 29.5 65.3 24.4
Tape Recorder 9.7 15.2 19.3 32.6 62.9 27.9
Television 4.5 10.5 18.1 36.9 75.6 29.1
Satellite dish 0.2 0.3 0.7 3.7 36.4 8.3
Video cassette recorder/DVD 0.6 1.4 3.4 10.4 47.3 12.6
Telephone/Cell phone 5.0 12.8 23.4 43.5 82.8 33.5
Refrigerator 4.0 9.4 19.8 38.9 79.5 30.3
Stove, gas or electric 10.4 20.0 34.1 59.9 88.2 42.5
Microwave 0.1 1.0 2.6 8.3 46.5 11.7
Freezer 1.4 4.6 9.9 21.9 58.4 19.3
Washing machine 0.9 1.9 4.8 11.9 49.8 13.9
Motor vehicle 1.5 3.6 8.1 18.6 60.6 18.5
Motor cycle/Scooter 0.2 0.4 0.1 0.5 4.2 1.1
Sewing/Knitting machine 9.6 12.1 13.4 15.7 28.7 15.9
Bicycle 8.7 14.4 14.2 15.5 25.4 15.6
Computer .. 0.1 0.3 1.7 25.0 5.4
Internet service .. 0.1 .. 0.2 13.7 2.8
Canoe/Boat 2.2 2.1 1.3 1.1 1.0 1.5
Motorboat .. 0.1 0.1 .. 1.1 0.3
Camera 1.5 3.4 6.9 13.0 44.3 13.8
Agricultural assets
Donkey cart/Ox cart 10.0 10.5 9.0 6.6 5.4 8.3
Plough 27.4 34.9 25.4 17.0 8.3 22.6
Tractor 0.4 0.1 0.4 0.8 4.6 1.3
Wheelbarrow 9.9 16.2 18.6 20.1 30.6 19.1
Grinding mill 0.2 1.0 1.1 1.8 5.1 1.9
Cattle 31.2 37.9 34.8 34.6 30.0 33.7
Sheep 3.5 4.3 5.9 6.9 11.6 6.4
Pig 16.6 23.1 17.6 10.4 3.7 14.3
Goat 40.5 48.0 41.9 37.9 26.8 39.0
Donkey/mule 19.9 23.4 19.6 14.1 9.5 17.3
Horse 4.2 3.7 4.1 5.6 9.4 5.4
Poultry 61.1 66.4 55.8 38.3 21.5 48.6
Ostrich 0.1 0.2 0.2 0.2 1.5 0.4
Grazing land 2.2 3.4 3.1 5.2 9.4 4.7
Field for crops 35.7 30.8 26.7 19.9 12.5 25.1
- 20 -
Table 8: Incidence of poverty by ownership/access to agricultural assets, 2003/2004
(%)
Severely poor Poor Non-poor Namibia
Owns or has access to cattle
Owns 29.7 32.4 34.2 33.7
Does not own, but has access 10.0 10.0 5.9 7.1
Neither owns nor has access 60.1 57.5 59.8 59.1
Not stated 0.1 0.1 0.1 0.1
Total 100.0 100.0 100.0 100.0
Owns or has access to goat
Owns 37.9 42.7 37.6 39.0
Does not own, but has access 2.6 2.9 3.9 3.6
Neither owns nor has access 59.2 54.1 58.4 57.2
Not stated 0.4 0.3 0.1 0.2
Total 100.0 100.0 100.0 100.0
Owns or has access to field for crops
Owns 35.4 34.7 21.5 25.1
Does not own, but has access 28.9 34.7 27.0 29.1
Neither owns nor has access 35.2 30.2 51.3 45.5
Not stated 0.5 0.4 0.2 0.3
Total 100.0 100.0 100.0 100.0
Ownership/access to plough
Owns 25.3 30.0 19.8 22.6
Does not own, but has access 18.5 18.8 10.7 13.0
Neither owns nor has access 55.8 50.9 69.1 64.1
Not stated 0.4 0.3 0.3 0.3
Total 100.0 100.0 100.0 100.0
Among all Namibian households, 71.4 percent own a radio while an additional 13.1 have
access to one (Table 9). Ownership is higher among the non-poor households where 75.3
percent own a radio, than in poor and severely poor households where 61.2 and 59.0 per-
cent, respectively claim ownership. Much more unequal is the ownership of telephones.
Among non-poor households, 44 percent own a telephone (including cell phones) com-
pared to 5.9 and 4.6 percent among poor and severely poor households, respectively.
Poor and non-poor households claim higher rates of access, rather than ownership, for
instance through borrowing or public phones. However, among both poor and severely
poor households, more than half, 53.7 and 57.7 percent respectively, neither own nor
have access to a telephone.
- 21 -
Table 9: Incidence of poverty by ownership/access to radio, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Owns 59.0 61.2 75.3 71.4
Does not own, but has access 19.6 19.0 10.9 13.1
Neither owns nor has access 21.3 19.6 13.6 15.3
Not Stated 0.1 0.2 0.1 0.1
Total 100.0 100.0 100.0 100.0
Table 10: Incidence of poverty by ownership/access to telephone/cell phone,
2003/2004 (%)
Severely poor Poor Non-poor Namibia
Owns 4.6 5.9 44.0 33.5
Does not own, but has access 37.4 40.4 30.5 33.3
Neither own nor has access 57.7 53.4 25.2 33.0
Not stated 0.4 0.4 0.2 0.3
Total 100.0 100.0 100.0 100.0
Figure 13: Share of households that own various assets, 2003/2004 (%)
0% 20% 40% 60% 80%
Computer
Internet service
Motor cycle/scooter
Satellite dish
VCR/DVD
Washing machine
Motor vehicle
Camera
Freezer
HiFi
Refrigerator
TV
Telephone/cellphone
Sewing/knitting machine
Bicycle
Tape recorder
Radio
Poor
Non-poor
- 22 -
Distance and access
The distinct geographical dimensions of poverty in Namibia are to some extent reflected in
the distance variables that are included in the NHIES. Figure 14 shows how poor house-
holds are generally farther away in distance measured in kilometres from a range of admin-
istrative and infrastructural services compared to non-poor households. The corresponding
national and regional figures are included in Table 13. For instance, the average distance of
a poor household to a magistrate court is 44.6 kilometres compared to 29.0 kilometres for a
non-poor household. Poor households are on average 31.0 kilometres away from a secon-
dary school compared to 23.9 kilometres for non-poor households. The distance to a police
station is 14.1 and 24.1 kilometres for non-poor and poor households, respectively. Among
all the facilities and services, the distance to drinking water is the lowest (i.e. the facility is
nearest to the household) for both groups but still the poor have more than twice the dis-
tance (1.1 kilometres) on average to access drinking water compared to non-poor house-
holds (0.5 kilometres).
Figure 14: Average distances to facilities and services, 2003/2004 (kilometres)
0 5 10 15 20 25 30 35 40 45 50
Drinking water
Traditional court
Primary school
Market/shop
Public transport
Hospital/clinic
Police station
Post office
Secondary school
Combined school
Magistrate court
Poor
Non-poor
There are discernable differences between regions in the distances of poor households to
services and facilities, which is a reflection of several factors including the availability of
infrastructure, population density and urbanisation. For instance, in Omaheke the average
distance to a hospital or clinic for a poor household is 30.2 kilometres, in Oshana the av-
erage distance is 5.7 kilometres. In Omaheke, the average distance for poor households to
public transportation is 22 kilometres whereas in Caprivi it is 2.2 kilometres. It should be
noted that the physical distance between the household and these facilities and services
are generally expected to have less adverse impacts in non-poor households as these are
- 23 -
more likely than poor households to own a motor vehicle or have access to one (see fur-
ther below), or have income available to incur public transportation costs.
Housing and utilities
Housing and utilities are major categories of household expenditure and thus key deter-
minants of the non-food component of the cost of basic needs poverty line. Moreover,
incidence of poverty is correlated with a series of physical housing characteristics and
utilities. Overall, 64.9 percent of households in Namibia owned their own house. Home
ownership is higher for poor and severely poor than for the non-poor. While 56.5 percent
of non-poor households own their home, the corresponding shares for poor and severely
poor households are 86.8 percent and 88.3 percent, respectively. This typically refers to
communal housing. The second most common type of tenure among poor and severely
poor households is Occupied free.
Table 11: Incidence of poverty by type of tenure, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Owned 88.3 86.8 56.5 64.9
Owned but not paid off 1.6 1.7 15.2 11.5
Occupied free 7.8 8.9 12.2 11.3
Rented w/o subsidy 2.2 2.5 13.6 10.5
Rented with subsidy 0.1 0.1 2.5 1.8
Namibia 100.0 100.0 100.0 100.0
Table 12 shows the correlation between poverty status and type of dwelling. Among non-
poor households, 43.1 percent live in a detached house and 4.3 percent in an apartment.
Together those two categories are often referred to as modern dwelling. The shares of
poor and severely poor households that reside in a modern dwelling are 10.3 and 8.9 per-
cent, respectively. The majority of poor and severely poor live in traditional dwellings
and a large share of both poor and non-poor live in improvised housing units (defined as
housing built with discarded materials such as in informal settlements).
Table 12: Incidence of poverty by type of dwelling unit, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Detached or semi-detached
house 8.6 10.0 43.1 34.0
Apartment/flat 0.3 0.3 4.3 3.2
Traditional dwelling 67.5 69.2 34.1 43.8
Improvised housing unit 22.2 19.2 15.6 16.6
Other 1.4 1.3 2.9 2.4
Namibia 100.0 100.0 100.0 100.0
- 24 -
Table 13: Average distances to facilities and services by region, 2003/2004 (kilometres)
Region
Drinking
water
Hospital/
clinic
Public
transport
Market/
shop
Primary
school
Secondary
school
Combined
school
Police
station
Post
office
Magistrate
court
Caprivi Non-poor 0.6 5.1 2.2 3.8 2.6 13.8 4.8 11.2 23.1 35.9
Poor 0.6 7.7 2.1 4.4 2.6 13.7 5.7 19.1 25.3 52.5
Erongo Non-poor 0.1 4.3 2.2 2.8 3.8 8.6 26.6 4.7 5.5 11.3
Poor 0.3 8.4 7.7 7.8 7.3 31.6 72.8 9.2 24.0 50.0
Hardap Non-poor 0.1 18.0 18.8 15.4 16.3 43.0 139.2 26.0 25.1 32.2
Poor 0.5 15.1 14.2 11.5 12.1 41.9 161.9 23.8 23.9 29.8
Karas Non-poor 0.1 16.6 12.3 11.0 12.1 76.1 123.5 16.3 15.8 45.0
Poor 0.2 17.2 15.6 12.5 12.0 56.5 65.1 21.1 19.5 49.0
Kavango Non-poor 1.3 5.4 3.9 3.7 2.0 17.8 10.4 23.9 26.7 28.7
Poor 1.9 7.7 7.3 7.1 4.6 24.1 13.8 42.3 48.4 50.6
Khomas Non-poor 0.0 5.2 2.2 3.1 3.9 6.1 7.3 5.4 5.8 7.6
Poor 0.1 11.3 7.3 7.4 9.5 17.9 20.2 11.6 15.4 17.7
Kunene Non-poor 0.7 32.8 23.1 21.3 17.2 62.9 75.1 36.6 48.8 51.9
Poor 0.7 21.4 14.6 18.0 13.2 50.0 74.2 29.9 36.8 43.5
Ohangwena Non-poor 1.5 12.4 6.2 10.3 4.0 24.8 5.7 17.4 34.3 40.8
Poor 1.3 10.0 8.2 7.6 3.4 23.8 4.9 17.7 32.2 36.0
Omaheke Non-poor 0.2 34.7 31.7 18.4 30.0 113.2 200.0 47.5 65.9 100.7
Poor 0.3 30.2 22.0 9.6 19.2 124.7 238.9 38.6 36.9 91.1
Omusati Non-poor 1.2 8.3 4.6 4.2 3.6 16.9 6.1 13.2 18.2 39.9
Poor 1.3 9.4 5.4 3.6 3.4 21.3 6.5 20.7 26.4 43.4
Oshana Non-poor 0.6 4.5 1.8 5.1 2.0 8.4 2.4 7.1 10.7 12.3
Poor 1.0 5.7 3.7 6.5 2.3 12.8 2.9 10.5 13.4 14.2
Oshikoto Non-poor 1.2 12.9 5.2 4.7 6.7 20.2 8.6 17.3 16.3 50.8
Poor 1.4 18.1 5.9 6.3 9.6 27.8 13.1 23.3 21.4 58.1
Otjozondjupa Non-poor 0.1 20.5 5.0 11.6 15.3 34.8 41.2 16.2 23.7 29.2
Poor 0.2 19.4 7.6 15.0 18.4 43.5 36.7 20.1 36.7 43.9
Namibia Non-poor 0.5 10.8 6.2 6.9 6.9 23.9 31.1 14.1 18.7 29.0
Poor 1.1 12.5 8.0 8.0 7.4 31.1 33.9 24.1 30.4 44.6
- 25 -
Poverty levels are also reflected in access to water and sanitation facilities. Table 14
shows poverty levels by a range of possible sources of drinking water. In Namibia, 28.6
percent of households have piped water in the dwelling, 25.7 percent use a public tap and
14.6 percent have access to piped water on the site of the dwelling. Among the non-poor
households, 37.9 percent have piped water in the dwelling compared to 4.2 and 3.3 per-
cent of poor and severely poor, respectively. The main source of drinking water for poor
households is public tab, which 36.4 percent of households rely on. Communal bore hole
is the main source of drinking water for 10.8 percent of poor households, 8.9 percent rely
on flowing water and 7.8 percent on unprotected wells.
Table 14: Incidence of poverty by source of drinking water, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Piped in dwelling 3.3 4.2 37.9 28.6
Piped on site 10.5 11.5 15.8 14.6
Neighbor's tap 7.5 7.5 4.6 5.4
Public tap 35.5 36.4 21.7 25.7
Water carrier or tanker 0.8 0.6 0.6 0.6
Private Bore Hole 4.1 3.9 1.8 2.4
Communal bore hole 11.4 10.8 5.4 6.9
Protected well 4.3 4.6 2.3 2.9
Spring 0.1 0.1 0.1 0.1
Flowing water 11.7 8.9 3.1 4.7
Rain Water Tank 0.5 0.5 0.1 0.2
Unprotected well 7.5 7.8 4.5 5.4
Dam/Pool/Stagnant water 2.4 2.8 1.6 1.9
Other 0.5 0.4 0.3 0.4
Total 100.0 100.0 100.0 100.0
Table 15: Incidence of poverty by sanitation facilities, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Flush/Sewer 5.7 7.3 44.5 34.3
Flush/Septic Tank 0.7 0.9 3.1 2.5
Pit Latrine/VIP 2.2 2.3 4.4 3.8
Pit Laterine/no ventilation 5.6 5.6 4.3 4.6
Bucket 2.3 1.9 1.0 1.3
Other 0.2 0.2 0.2 0.2
Bush 83.4 81.8 42.5 53.3
Total 100.0 100.0 100.0 100.0
A similar picture emerges when comparing the incidence of poverty by sanitation facili-
ties. In Namibia as a whole, 34.3 percent of households have a flush/sewer sanitation sys-
tem compared to 7.3 percent among the poor and 5.7 percent among the severely poor.
More than 80 percent of poor and severely poor households use the bush as a toilet,
which is almost double the rate for non-poor households. More than half of all Namibian
households, 53.3 percent, rely on the bush as the main toilet facility. Less than 4 percent
- 26 -
of all households in Namibia use a ventilated improved pit. Among the non-poor, 1.0
percent use a bucket compared to 1.9 and 2.3 percent among the poor and severely poor,
respectively.
The housing quality is measured by the material for roof, wall and floors, and poor
households stand out on all these variables. For instance, just over 50 percent of poor and
severely poor households have thatched roofs, more than double the share among non-
poor households (Table 16). Among non-poor households, 62 percent use iron or zinc
compared to 35.8 and 37.1 percent among poor and severely poor households, respec-
tively. Similarly, while 49 percent of non-poor households have their house walls built
from cement blocks, the shares among poor and severely poor households are 13.3 and
10.4 percent, respectively (Table 17). More than 38 percent of poor and severely poor
households use either sticks, mud, clay or dung. When it comes to the material used for
the floor of the house, 58.4 percent of non-poor use concrete, while 54.7 and 57.0 percent
of poor and severely poor households, respectively use sand (Table 18).
Table 16: Incidence of poverty by material for roof, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Cement blocks 0.1 0.2 0.8 0.7
Bricks 0.2 0.2 0.3 0.3
Iron/Zinc 37.1 35.8 62.0 54.8
Poles/sticks/grass 9.7 10.2 5.8 7.0
Sticks/mud/clay/dung 0.8 0.6 0.6 0.6
Asbestos 0.4 0.9 6.1 4.6
Tiles .. .. 0.3 0.2
Slate 0.2 0.1 0.3 0.2
Thatch 50.2 50.9 21.9 29.9
Other 1.3 1.1 1.9 1.7
Total 100.0 100.0 100.0 100.0
Table 17: Incidence of poverty by material for the wall, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Cement blocks 10.4 13.3 49.0 39.1
Bricks 1.4 1.6 3.1 2.7
Iron/Zinc 22.0 18.6 14.1 15.4
Poles/sticks/grass 20.4 22.3 12.1 14.9
Sticks/mud/clay/dung 38.9 38.8 17.3 23.2
Asbestos 0.2 0.4 0.7 0.6
Tiles 0.2 0.1 0.3 0.2
Slate .. 0.0 0.1 0.1
Thatch 3.5 2.4 0.9 1.3
Other 2.9 2.4 2.5 2.4
Total 100.0 100.0 100.0 100.0
- 27 -
Poverty status and energy access are closely correlated. Among poor and severely poor
households, 89.7 and 91.6 percent, respectively depend on wood as an energy source for
cooking (Table 19). Among non-poor households, 38 percent depend on electricity from
the main grid; more than ten times the share among poor households. Only 5.8 percent of
households in Namibia use gas for cooking and in total more than half, 59.6 percent of all
households, poor and non-poor, rely on wood as their main source of energy for cooking.
When it comes to energy for lighting, the main source among poor and severely poor
household is candles, 54.6 and 56.0 percent, respectively (Table 20). For 46.2 percent of
non-poor households the main source of energy for lighting is the main grid.
Wood remains the most used source of heating energy for Namibian households at 45.8
percent, but with much higher shares among the poor and severely poor households, 66.1
and 64.3 percent, respectively (Table 21). Nearly one third of households in Namibia do
not use any energy for heating and the shares are only slightly lower among poor and se-
verely poor households compared to non-poor households. It is interesting to note that
even if 8.5 percent of households source their electricity for lighting from the main grid,
less than half use the main grid for cooking and even fewer for heating. This is an indica-
tion that poor households switch between energy sources depending on purpose. It is also
noteworthy that less than one percent of households, irrespective of poverty status, use
solar energy for either cooking, heating or lighting.
Table 18: Incidence of poverty by material for the floor, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Sand 57.0 54.7 28.9 36.0
Concrete 17.6 19.5 58.4 47.7
Mud/clay/and/or dung 24.8 25.3 11.3 15.2
Wood 0.0 0.2 0.6 0.5
Other 0.4 0.3 0.7 0.6
Not stated 0.1 0.0 0.0 0.0
Total 100.0 100.0 100.0 100.0
Table 19: Incidence of poverty by energy source for cooking, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Electricity from mains 2.6 3.6 38.0 28.5
Electricity from generator 0.1 0.4 0.3
Solar Energy .. 0.0 0.0 0.0
Gas 1.7 2.3 7.2 5.8
Paraffin 2.1 2.6 5.1 4.4
Wood 91.6 89.7 48.1 59.6
Coal 0.2 0.2 0.1 0.2
Animal Dung 1.7 1.3 0.9 1.0
Other .. 0.1 0.0 0.0
None .. 0.1 0.1 0.1
Total 100.0 100.0 100.0 100.0
- 28 -
Table 20: Incidence by energy source for lighting, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Electricity from mains 7.3 8.5 46.2 35.8
Electricity from generator 0.2 0.2 0.9 0.7
Solar Energy 0.1 0.1 0.6 0.4
Gas 0.1 0.1 0.2 0.2
Paraffin 13.7 16.8 14.5 15.1
Wood 15.9 14.2 3.4 6.4
Candles 56.0 54.6 32.2 38.4
Other 5.6 4.6 1.7 2.5
None 1.1 1.0 0.3 0.5
Total 100.0 100.0 100.0 100.0
Table 21: Incidence by energy source for heating, 2003/2004 (%)
Severely poor Poor Non-poor Namibia
Electricity from mains 1.4 1.6 24.8 18.4
Electricity from generator 0.1 0.0 0.3 0.3
Solar Energy .. 0.1 0.1 0.1
Gas 0.1 0.1 1.0 0.7
Paraffin 0.5 0.3 1.0 0.8
Wood 64.3 66.1 38.1 45.8
Coal 1.3 1.0 0.6 0.7
Animal Dung 1.1 0.8 0.7 0.7
Candles 0.4 0.3 0.3 0.3
Other 0.2 0.4 0.3 0.3
None 30.7 29.2 32.9 31.9
Total 100.0 100.0 100.0 100.0
- 29 -
5. Determinants of consumption and poverty
The previous sections of this report have highlighted a number of features that character-
ise the poor, severely poor and non-poor households. While this is useful in describing
how each variable correlates one by one to the poverty status of households, such an
analysis can oversimplify complex relationships. Multivariate analysis on the other hand
makes it possible to determine the effects that accrue from each variable when simultane-
ously controlling for the effect of all others. This way it is possible to gauge, for instance,
whether the observed differences between households in urban and rural areas are spe-
cific to location or whether differences are more attributable to variation in other charac-
teristics of urban and rural households such as educational attainment, household compo-
sition and source of income. This section briefly highlights the results from two types of
multivariate analysis; first on the determinants of household consumption expenditure
and second on the poverty status of households. More details on the methodology and
more in-depth results are in Annex G. Once again it should be noted that the effects here
relate to correlation and that no aspects of causation can be inferred.
Determinants of household poverty status
The main findings of the first multivariate analysis are summarised in Figure 15. The fig-
ure shows an inverse relationship between household expenditure and the size of the
household. Increasing the size of the household (by one adult equivalent) reduces total
household expenditure by 23.9 percent when all other factors are controlled for. Female-
headed households have total consumption expenditures that are lower by 4.9 percent
compared to male-headed households. As expected, given the results of the poverty pro-
file, household consumption expenditure increases with the age of the head of household.
Moreover, having one or more children in the household reduces adult equivalent con-
sumption by 12 percent compared to households without any children and holding other
factors, including household size, constant.
The analysis confirms the great regional differences in levels of consumption expenditure
among households. Rural households also have lower levels of consumption expenditure
compared to the urban default controlling for all other factors. In households where the
head has primary education as the highest level of education or has no formal education
at all, the monthly consumption levels are lower by 19.8 and 24.4 percent, respectively
compared to households where the head has attained a secondary level of education.
Conversely, in households where the head has attained a tertiary education, the consump-
tion levels are higher by 26.6 percent compared to household heads with a secondary
education. Having a pension as the main source of income reduces consumption expendi-
ture by 4.6 percent compared to all other sources of income including wages, income
from subsistence farming and non-farming business activities. The variables reflecting
distances to public services and facilities are somewhat ambiguous. Expenditure levels
increase with distance to hospital/clinic and shop/market but decrease with distance to
police station.
Households of Caprivi and Kavango have lower levels of household consumption when
controlling for other factors. Also, the regions of Karas, Hardap and Oshikoto have lower
- 30 -
levels of consumption expenditure in comparison with Ohangwena as the default cate-
gory. On the other hand, Khomas, Omusati and Oshana have higher levels of consump-
tion expenditure. This may seem to differ from the results from the poverty profile, which
showed that Ohangwena ranked second highest in terms of both levels of poverty and
poverty share. The reason for the change in ranking is that the multivariate analysis con-
trols for other factors that determine poverty status and shows the strength of the effects
that are attributable to the region per se. This way, the results show that when holding
constant all other characteristics that are thought to influence income and consumption
levels e.g. education levels, age, number of children in the household and so on, a house-
hold in Caprivi is likely to be poorer than a household living in any other region of the
country. Likewise, a household in Khomas is more likely to have a higher level of in-
come or consumption than in any other region.
Figure 15: Determinants of household consumption expenditure (percentage change)
-30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0%
Household size
Age of head of household
Age of head of household (squared)
Female
Child younger than 16
Widow/widower
Rural
Distance to hospital/clinic (km)
Distance to police station (km)
Distance to shop/market (km)
Owns or has access to field for grazing
Pension
Primary education
Tertiary education
No formal education
Caprivi
Erongo
Hardap
Karas
Kavango
Khomas
Omusati
Oshana
Oshikoto
Note: The table shows the results from the OLS regression on log of total monthly adult equivalent expendi-
ture. Only results significant at 10% or lower are reported. The regression also included dummy variables
for language groups, which are not shown but reported in Annex G along with the full regression output.
The omitted categories for the categorical variables are; male, no child younger that 16, marital status
other than widow/widower, urban, neither owns nor has access to field for crops, other income sources,
secondary education, Ohangwena.
In households where Afrikaans is the main language, total consumption is higher by 19.8
percent compared to the default category, which is Oshiwambo, and households where
German and English consumption is higher by 11.3 and 10.5 percent, respectively con-
trolling for other factors. On the other hand, households where the main languages spo-
- 31 -
ken are Khoisan, Rukavango and especially Nama/Damara total consumption levels are
lower (than the default category), again holding constant all other factors.
Determinants of household poverty status
The second type of multivariate analysis conducted on the data makes use of the new
poverty line definition by predicting the probability or the odds ratio of a household being
poor given the range of background variables. Results are reported in Figure 16 where the
odds ratios have been ranked from highest to lowest for illustration purposes. The higher
the odds ratio, the higher the probability that the household will be poor. The highest
odds ratio is for no formal education of the head of household. These households have an
odds ratio of 4.2. In other words, households where the head has no formal education are
more than four times as likely to be classified as poor compared to households where the
head has a secondary education and controlling for all other factors. Households where
primary education is the highest level of education attained by the head of household are
also more likely to be poor. The analysis further shows that households in rural areas
have an odds ratio of 1.97, which means that they are 97 percent more (almost twice as)
likely to be poor compared to urban households and holding all other factors constant.
Figure 16: Probabilities of households being poor (odds ratios)
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Tertiary education
Khomas
Oshana
Kunene
Erongo
Owns or has access to field for grazing
Omusati
Widow/widower
Owns or has access to field for crops
Age of head of household
Distance to hospital/clinic (km)
Age of head of household (squared)
Distance to police station (km)
Female
Household size
Oshikoto
Kavango
Pension
Child younger than 16
Rural
Caprivi
Primary education
No formal education
Note: The table shows the results from the binary logistic regression on poverty status (poor=1 and non-
poor=0). Only results significant at 10% or lower are reported. The regression also included dummy vari-
ables for language groups, which are not shown but reported in Annex G along with the full regression
output. The omitted categories for the categorical variables are; male, no child younger that 16, marital
status other than widow/widower, urban, neither owns nor has access to field for crops, other income
sources, secondary education, Ohangwena.
- 32 -
Additional factors contribute to the probability of household poverty. Having a child
younger than 16 in the household make it 1.77 times (or 77 percent) more likely to be
poor compared to households without any children. Households where pension is the
main source of income are 1.74 times more likely to be poor than households that rely on
other main sources of income. Female-headed household are 1.18 times as likely to be
poor compared to male-headed households. Several regional variables, Caprivi, Kavango
and Oshikoto, also have odds ratios higher than one, which indicates that households re-
siding in these regions are more likely to be poor, compared to households residing in
Ohangwena (the default category) and holding all other variables constant.
Conversely, several factors have odds ratios below 1, which means that the probabilities
shift towards the household being less likely than the default category to be classified as
poor. The most important of these factors is tertiary education. An odds ratio of 0.019
implies that if the household head has a tertiary education, it is 50 times less likely to be
poor compared to a household where the head has a secondary education. Moreover,
households residing in the regions of Erongo, Kunene, Oshana and Khomas are half as
likely to be poor compared to those in Ohangwena when all other factors are controlled
for.
- 33 -
6. Inequality and polarisation
As explored throughout the preceding sections of this report, consumption expenditure and
levels of poverty are distributed very unevenly in Namibia. Since the promotion of social
equity and the reduction of inequality remain a high priority for national development pol-
icy, it is important to establish measures that can adequately reflect levels of inequality,
proximate factors and drivers of change. In this section, the analysis of poverty is supple-
mented by additional measures on inequality and polarisation. The section is kept brief and
non-technical in line with the desire to make the analysis presented in this report accessible
to as wide an audience as possible. A deeper and more technical analysis is included in
Annex H. The importance of the analysis presented in this section is underscored by the
general conclusion that Namibian society remains among the most unequal and polarised in
the world.
Table 22: Adult equivalent expenditure by deciles, 2003/2004
Distribution of household expenditure
Table 22 shows the distribution of expenditure by deciles, i.e. grouping together house-
holds in 10 equal size groups ranked by expenditure with decile 1 comprising households
with the lowest expenditure and decile 10 with the highest expenditure. The results show
that among those households with the 10 percent lowest monthly expenditure, the average
expenditure is N$ 116.20 per adult equivalent and the combined expenditure of this group
makes up 1.07 percent of total expenditure among all households. At the other end of the
distribution, among the top 10 percent, average monthly expenditure is N$ 5743.88. The
table also shows the percentage share that each decile claims out of total expenditure. Ex-
penditures in the top decile, even if only including 10 percent of households, constitute
more than half, 53 percent, of total expenditure of all households. At the other extreme, ex-
penditure among the lowest decile makes up just over 1 percent of total expenditure of all
households. For all deciles, average expenditure are lower among rural households com-
pared to urban ones, which is expected given the results from the poverty profile. The aver-
Mean expenditure (N$) Share of total expenditure (%)
Decile Urban Rural Total Urban Rural Total
1 161.44 103.50 116.20 0.95 1.57 1.07
2 298.22 167.01 191.79 1.75 2.54 1.77
3 415.59 209.16 247.24 2.43 3.17 2.28
4 562.08 247.98 311.67 3.30 3.76 2.88
5 730.38 296.38 387.42 4.29 4.50 3.58
6 961.35 352.94 500.22 5.62 5.36 4.61
7 1312.41 426.72 673.67 7.70 6.46 6.23
8 1903.56 557.38 968.62 11.15 8.46 8.93
9 3241.26 805.04 1691.93 19.04 12.20 15.64
10 7481.81 3419.57 5743.88 43.77 51.99 53.01
Total 1705.76 659.14 1083.03 100.00 100.00 100.00
- 34 -
age adult equivalent expenditure in rural households is almost one third of expenditure in
urban households. Figure 17 compares the share of total expenditure by each decile with
the corresponding share of the total population. While each decile contains the same num-
ber of households, because poorer households tend to have more members, population
shares at the lower end exceed 10 percent. By focusing on individuals rather than house-
holds gives an even more disturbing picture of inequality in Namibia. While the 40 percent
of households with the lowest expenditure hold more than half the population (51.97 per-
cent), their total expenditure comes to just 8 percent of total expenditure in the country.
Moreover, the 10 percent of richest households are home to just 5.62 percent of the popula-
tion but these command more than half (53.01) of total household expenditure.
Figure 17: Share of total expenditure and share of total population by household dec-
ile, 2003/2004
0%
10%
20%
30%
40%
50%
60%
Households by decile
Expenditure share
Population share
Expenditure share 1.07% 1.77% 2.28% 2.88% 3.58% 4.61% 6.23% 8.93% 15.64% 53.01%
Population share 14.96% 13.47% 12.28% 11.26% 10.27% 9.15% 8.42% 7.54% 7.04% 5.62%
1 2 3 4 5 6 7 8 9 10
The Lorenz curve and the Gini index
Another popular way of expressing inequality graphically is through a Lorenz diagram,
which plots the cumulative share of consumption expenditure against the cumulative share
of households. A 45 degree line represents a situation whereby the cumulative share of
households and their cumulative consumption are the same at all levels thus indicating the
hypothetical situation whereby there is perfect inequality in the way expenditures are dis-
tributed (everyone gets the same). The further away the observed Lorenz curve is from the
45 degree curve, the greater is the inequality in the distribution. Figure 18 presents the Lo-
renz curves for all households, and for rural and urban ones separately. Two issues emerge
from the figure. Firstly, note how at all points the line for urban households is to the left of
the line representing the total number of households. This implies that as a group inequality
- 35 -
among urban households is lower than for all households together. Secondly, the line for
rural households intersect both lines for urban and all households. This implies that no firm
conclusion about the comparative levels of overall inequality between urban and rural areas
can be made on the basis of a visual inspection of the Lorenz diagram alone. Other meas-
ures need to be applied.
Figure 18: Lorenz diagram, 2003/2004
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Cumulative share of households
C
u
m
u
la
ti
v
e
s
h
a
re
o
f
e
x
p
e
n
d
it
u
re Total
Urban
Rural
One such measure is the Gini coefficient, which is computed as the distance between the
Lorenz curve and the 45 degree line, and provides a numerical value of the degree of ine-
quality. The Gini-coefficient takes a value between 0 and 1, where 0 represents the unlikely
situation of perfect equality where all households have the exactly the same level of con-
sumption expenditure. A value of 1 for the Gini-coefficient represents the equally unlikely
situation of the most extreme inequality whereby one household commands all the con-
sumption expenditure.
Table 23 provides an overview of Gini coefficients across a range of social and demo-
graphic variables. The table shows that the Gini-coefficient in 2003/2004 was 0.63. The
coefficient for both urban and rural areas is 0.58 indicating that inequality is similar across
localities. Naturally that does not imply that poverty levels are equal; as discussed above
poverty levels are much higher in rural areas. The fact that the Gini-coefficients for urban
and rural areas are lower than the national average is a further indication that lower in-
comes are concentrated in rural areas and higher incomes are concentrated in urban areas.
There are great differences in the degree of inequality in the 13 administrative regions of
Namibia. The lowest Gini-coefficients are found in Ohangwena and Omusati at 0.45 and
- 36 -
0.46, respectively. The highest is in Hardap with 0.69 and Omaheke with 0.64. Inequality is
lower in most regions but in Hardap and Kavango, the Gini-coefficients have moved up
significantly. The Gini-coefficient for those with no formal education is 0.39 suggesting
lower inequality among this group compared to those with any other level of education.
This is an indication that among those with no formal education, most have low incomes.
Similarly, in households were subsistence farming is the main source of income, the ine-
quality measure is lower than for other income sources, indicating a uniformity of low in-
comes in this category. For business and pension income, the Gini-coefficients are much
higher, which reflects a greater diversity of income levels, and thus higher inequality, in
these categories. Inequality is also higher in male-headed households compared to female-
headed households, and the levels of inequality generally increase with the age of the head
of household. Among households where English and German are the main languages spo-
ken, inequality is the lowest. These are also the households with the highest incomes.
Table 23: Gini-coefficients of households by social and demographic variable,
2003/2004
Namibia 0.63 Female 0.58
Male 0.64
Urban 0.58
Rural 0.58 16-20 0.45
21-24 0.49
Caprivi 0.47 25-29 0.59
Erongo 0.57 30-34 0.60
Hardap 0.69 35-39 0.58
Karas 0.61 40-44 0.62
Kavango 0.55 45-49 0.62
Khomas 0.57 50-54 0.68
Kunene 0.51 55-59 0.64
Ohangwena 0.45 60-64 0.70
Omaheke 0.64 65+ 0.60
Omusati 0.46
Oshana 0.56 Khoisan 0.44
Oshikoto 0.51 Caprivi languages 0.49
Otjozondjupa 0.60 Otjiherero 0.53
Rukavango 0.51
Primary education 0.43 Nama/Damara 0.52
Secondary education 0.55 Oshiwambo 0.52
Tertiary education 0.47 Setswana 0.50
No formal education 0.39 Afrikaans 0.56
German 0.31
Salaries/Wages 0.58 English 0.41
Subsistence Farming 0.38
Commercial Farming 0.52
Business 0.67
Pensions 0.66
On Figure 19, the Gini-coefficients of selected countries are presented. The figure is com-
piled using country data for the most recent survey where the income definition is house-
- 37 -
hold consumption, the sample is for full national coverage and the household is the unit of
analysis. However, since the data sources are individual country surveys where methodolo-
gies invariably differ, comparisons should be made with caution. Nevertheless, the figure
gives an indication of how Namibia fares globally. The Gini-coefficient for Namibia makes
the country rank high among the most unequal societies in the world when it comes to the
distribution of incomes (as measured by household consumption expenditure).
Figure 19: Gini-coefficients for selected countries
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Denmark
Sweden
Russian Federation
Korea, Republic of
Germany
Bangladesh
Canada
Indonesia
United Kingdom
Tanzania
Mauritius
Mozambique
Ghana
Kenya
Thailand
Cambodia
Cote d`Ivoire
China
United States
Singapore
Malawi
Argentina
Nigeria
Botswana
Chile
Zambia
South Africa
Brazil
Haiti
Namibia
Source: NHIES 2003/2004 for Namibia. For all other countries; the World Income Inequality Database of UN
World Institute for Development Economics Research.
Several explanations have been offered for the extreme levels of inequality in Namibia
(United Nations 2005). Notably, the countrys system of Apartheid rule prior to Independ-
ence in 1990, which was founded on policies of racial division and severely restricting ac-
cess to economic and social resources for the majority. Moreover, the countrys traditional
heavy reliance on extraction of natural resources, e.g. diamonds, means that production in
Namibia is highly intensive in the use of capital rather than labour. Note also how Figure
19 reveals that high levels of inequality is a particular challenge for counties in southern
Africa and Latin America, while at the other end of the inequality spectrum, countries in
Europe and especially in Scandinavia have low levels of income inequality.
Additional measures of inequality and polarisation
A series of additional measures of inequality and polarisation are presented in more detail
in Annex H. An important conclusion from this analysis is that inequality in Namibia is a
- 38 -
product not so much of differences between various population sub-groups as it is of differ-
ences within the same sub-groups. For instance, when decomposing inequality by the sex
of the head of household, it is shown that almost all of the prevailing inequality can be at-
tributed to inequalities within the two groups of male- and female-headed households and
much less to inequality between the two groups. Moreover, regional inequality is a result
more of inequalities within the regions and less so between them. This suggests that intra-
regional transfers are even more important in addressing inequality than inter-regional
transfers. The two sub-groups where between-group inequality is highestalthough the
within component also dominates hereare for education and language. This is a strong
indication that a large part of the inequality that exists in Namibia is attributable to differ-
ences in education levels and differences between language groups. This suggests that pub-
lic policy initiatives such as social transfers and empowerment initiatives need to be con-
cerned with both between and within types of differences and that targeting mechanisms
based on education and language would contribute substantially to reducing inequality.
The conventional inequality measures such as the Lorenz curve and the Gini-coefficient
may not be able to register important changes in the income distribution. The concept and
measures of polarisation seek to address this. Polarisation may be seen as a movement
from the middle of the income distribution towards the two tails leaving a hollowing of
the middle of the distribution. Two polarisation indices are calculated for Namibia in An-
nex H. The first measure follows Wolfson (1994), assumes two groups of equal size and
like the Gini index, is between 0 (no polarization) and 1 (complete polarization). The sec-
ond polarisation measure computed for the report is the Duclos-Esteban-Ray (DER) index,
which allows for individuals not to be clustered around discrete income intervals and
avoids arbitrary choices in the number of income groups through the use of non-parametric
estimation techniques (Duclos et al 2004). The results suggest that not only is Namibia one
of the most unequal societies in the world when it comes to income distribution, it also ap-
pears to be among the most polarised. For both indices, the values are higher in urban areas
than in rural areas indicating that polarization is greater in urban areas. Measures of polari-
sation as well as a broader range of inequality indicators as presented above could be added
to the indicators in the national poverty monitoring system to track developments over
time.
- 39 -
7. Conclusion
This report has presented an analysis of poverty and inequality in Namibia based on the
expenditure data from the 2003/2004 Namibia Household Income and Expenditure Survey
(NHIES) conducted by the Central Bureau of Statistics. The main innovation of the report
was the establishment of a new set of poverty lines for Namibia based on the Cost of Basic
Needs (CBN) approach, which has become part of the poverty monitoring standard in
SADC and most developing countries. Such poverty lines are particularly useful for draw-
ing of poverty profiles, examining the determinants of poverty and guiding policy interven-
tions aimed at poverty reduction.
Using the new CBN-based poverty lines, the study presented a detailed poverty profile of
Namibia. This profile showed that poverty status in the country is closely correlated with a
series of social, demographic, geographic and economic features of households. Multivari-
ate analysis confirms that poverty levels in Namibia are higher for instance among house-
holds that are female-headed, based in rural areas and have one or more children, when
controlling for other possible determinants. These results underscore the potential for pov-
erty reduction by greater targeting of policies and interventions. The report has shown dis-
cernible differences in the levels of consumption expenditure according to the education
levels of the head of household. These results underscore the centrality of the strengthening
the education system as an integral part of the national poverty reduction strategy. The re-
port also introduced a series of inequality measures beyond those traditionally applied in
Namibia. This part of the analysis re-affirmed that Namibia ranks among the most unequal
and polarised of societies in the world. Moreover, a decomposition exercise showed how
inequality is primarily a product of inequality within different population groups rather
than between these groups.
This report has also highlighted a range of methodological aspects in the establishment of
the poverty line for Namibia and documented the technical steps involved. However, the
analysis presented in this report must not be regarded in isolation but as part of a broader
effort that relies on quantitative as well as qualitative approaches to contribute to the under-
standing of poverty in Namibia as an important basis for designing effective interventions
to improve the welfare of Namibians.
- 40 -
References
Central Bureau of Statistics (CBS) 2006 The 2003/2004 Namibia Household Income and Expendi-
ture Survey National Planning Commission,Windhoek.
CBS 2001 Levels of Living SurveyNational Planning Commission, Windhoek.
CBS 1996 Living Conditions in Namibia: The 1993/1994 Namibia Household Income and Expendi-
ture Survey National Planning Commission, Windhoek.
CBS 1995 The distribution of economic resources in the population of Namibia: Some highlights
National Planning Commission, Windhoek.
Deaton A and C Paxson 1998 Economies of Scale, Household Size, and the Demand for Food
Journal of Political Economy, 106(5):897-930.
Deaton A, 1997 The Analysis of Household Surveys: a Microeconometric Approach to Develop-
ment Policy Washington D.C. and Baltimore: The World Bank and Johns Hopkins University
Press.
Duclos JY and A Araar 2006 Poverty and Equity: Measurement, Policy and Estimation with DAD
Kluwer Academic Publishers.
Duclos JY, J Esteban and D Ray 2004 Polarization: Concepts, Measurements, Estimation Econo-
metrica 72(6):1737-1772.
Ekström, E 1998 Income Distribution and Labour Market Discrimination: A Case Study of Na-
mibia Research Institute of Industrial Economics Working Paper Series No. 502.
FAO/WHO/UNU 1985 Energy and Protein Requirements. Report of a Joint Expert Consultation
WHO Technical Report Series No 724, Geneva.
Foster JE, J Greer and E Thorbecke 1984 A class of decomposable poverty measures Econometrica
52(3):761-766.
Foster J and A Shorrocks 1988 Poverty orderings and welfare dominance Social Choice and Wel-
fare, 5(2-3):179-198.
Government of the Republic of Namibia (GRN) 2005 Poverty Monitoring Strategy National
Planning Commission, Windhoek.
GRN 2004 Vision 2030 Office of the President, Windhoek.
GRN 2002 National Poverty Reduction Action Programme, 2001-2005 National Planning Com-
mission, Windhoek.
GRN 2001 Second National Development Plan, 2001-2006 (NDP2) National Planning Commis-
sion, Windhoek.
GRN 1998 Poverty Reduction Strategy for Namibia National Planning Commission, Windhoek.
Lanjouw JO 2001 Demystifying poverty lines
http://www.undp.org/poverty/publications/pov_red/Demystifying_Poverty_Lines.pdf
Lanjouw P and M Ravallion 1995 Poverty and Household Size Economic Journal, 105(433):
1415-34.
Levine S 2007 Trends in Human Development and Human Poverty in Namibia United Nations De-
velopment Programme, Windhoek.
- 41 -
May J and B Roberts 2005 Poverty Diagnostics Using Poor Data: Strengthening the Evidence
Base for Pro-poor Policy Making in Lesotho Social Indicators Research, 74:477-510.
Ministry of Health and Social Services 2001 First Report of the Working Group on HIV/AIDS
Impact Projections for Namibia GRN, Windhoek.
Ravallion M 1998 Poverty Lines in Theory and Practice LSMS Working Paper No 133 World
Bank, Washington DC.
Ravallion M 1992 Poverty Comparisons: A Guide to Concepts and Methods LSMS Working Pa-
per No. 88 World Bank, Washington, DC.
Ravallion M and B Bidani 1994 How Robust is a Poverty Profile? World Bank Economic Review,
8(1):75102.
Ravallion M. and M Huppi 1991 Measuring Changes in Poverty: A Methodological Case Study of
Indonesia During an Adjustment Period World Bank Economic Review, 5:57-84.
Tarp F, K Simmler, C Matusse, R Heltberg and G Dava 2002 The Robustness of Poverty Lines
Reconsidered Economic Development and Cultural Change 51(1):77-108.
United Nations 2003 Indicators for Monitoring the Millennium Development Goals; Definitions,
Rationale, Concepts and Sources United Nations Statistics Division, New York.
United Nations 2005 Common Country Assessment for Namibia, Windhoek.
United Nations Statistics Division 2005 Handbook on Poverty Statistics: Concepts, Methods and
Policy Use New York.
Van Rooy G, B Roberts, C Schier, J Swartz, and S Levine 2006 Poverty and Inequality in Namibia
Multi-Disciplinary Research and Consultancy Centre Discussion Paper No 1.
Van Rooy G, G !Naruseb, M Maasdorp, G Eele, J Hoddinot, and S Stone 1994 Household Subsis-
tence Levels: Three Selected Communities in Namibia Social Science Division Research Report No
9.
White H and E Masset 2003 The Importance of Household Size and Composition in Constructing
Poverty Profiles: An Illustration from Vietnam Development and Change 34(1):105-126.
Wolfson M 1994 When Inequality Diverges American Economic Review 84:353-358.
World Bank 2007 World Development Indicators CD ROM, Washington DC.
Yaron G, G Janssen and U Maamberua 1992 Rural Development in the Okavango Region: an As-
sessment of Needs, Opportunities and Constraints NISER, Windhoek.
Zimmerman C 1932 Ernst Engel's Law of Expenditures for Food The Quarterly Journal of Eco-
nomics 47(1):78-101.
- 42 -
ANNEXES
ANNEX A: Background to the NHIES data
ANNEX B: Poverty measures in Namibia and SADC
ANNEX C: Setting the poverty line
ANNEX D: The national food basket
ANNEX E: First-order stochastic dominance tests
ANNEX F: Poverty profile tables
ANNEX G: Multivariate analysis
ANNEX H: Measures of inequality and polarisation
ANNEX I: Confidence intervals
- 43 -
ANNEX A: Background to the NHIES data
The data used for analysis in this report comes from the Namibia Household Income and
Expenditure Survey (NHIES) conducted in 2003/2004. The main results of the survey have
been published separately (CBS 2006) primarily in tabular form and with limited analysis
and interpretation. Since the official release of the NHIES results, the 2003/2004 data has
undergone further cleaning and refining for purposes of this report. This has not resulted in
major shifts in the data but in light even of small changes, caution should be exercised
when comparing the analysis in this report with previous releases of the data set. This An-
nex briefly describes the NHIES survey instrument. Some similarities and differences with
the first NHIES conducted in 1993/1994 are also highlighted. Due to methodological dif-
ferences between the two surveys, additional analytical work is ongoing to establish com-
parability, which will facilitate additional analysis on the dynamic aspects of poverty and
inequality in Namibia.
Table A-1: Key features of the NHIES
1993/1994 NHIES 2003/2004 NHIES
Dates of field work November 1993 to October
1994 (Walvis Bay was in-
cluded from May 1994)
1 September 2003 to 29
August 2004
Publication date
Sample size
October 1994 (preliminary)
May 1996 (full)
4,752 households
March 2006 (preliminary)
February 2007 (full)
10,920 households
Primary Sampling
Units
Response rate
192 (111 rural)
92.5%
546 (300 rural)
90%
Reference period for
the Daily Record Book
Four weeks Four weeks
Sources: CBS (2006, 1996)
About the surveys
The target population of the NHIES was the private household population of Namibia.
Therefore, the population living in institutions, such as hospitals, hostels, police barracks
and prisons were not covered. These were included in the Census, for instance. However,
households residing in private quarters of institutional settings were included. Some key
features of the two NHIES are listed in Table A-1. Two questionnaire forms were adminis-
tered to collect the data from the participating households. Form I was used to collect basic
information about the household and the people living in it, including: age, sex, education
- 44 -
and so on. Information on household incomes and expenditure were also collected on this
form using a 12 month reference period.
Form II, the Daily Record Book, was designed for households to record all expenditures
and receipts, item by item and including incomes and gifts (both received and given out),
every day. Each household would record these transactions daily over a four-week period.
In addition, households would record their consumption of goods from own-production, for
instance cereals, vegetables or eggs. These records detail the consumption of food and non-
food, as well as the flows of monetary and in-kind resources in and out of households all
over the country.
The survey was carried out over 13 of such four-week cycles each with a new set of house-
holds and thus a key distinguishing factor of the NHIES compared to other surveys con-
ducted by the Central Bureau of Statistics was that the NHIES was conducted over a full 12
months period. This ensured that any effects attributable to seasonality were evened out
and by changing the households every four weeks, respondent fatigue was minimised. It
is the detailed account of the consumption, incomes and expenditures of households gener-
ated by the survey that makes it so suitable for an analysis of poverty and material well-
being, which is the subject of this report.
Field work organisation
Recruitment of survey personnel was restricted to holders of Grade 12 Certificate or
equivalent. In the case of regional supervisors, a first degree was the minimum require-
ment. Advertisements were placed in newspapers and on various radio programmes for in-
terested persons to apply. Many applications were received from which suitable candidates
were selected. Academic qualification and previous survey experience were taken as crite-
ria for recruitment. Selection of the final core of field staff for each region was made from
regional trainees only. Only those who met recruitment requirements were selected from
the applicants for interviews. Due to large numbers of applicants, selection was by written
test. Efforts were made to recruit all language groups in each region to facilitate interviews
in local languages. A test was administered and those who passed were taken for deploy-
ment. A larger number of trainees, than the required compliment, were selected from appli-
cations from each region. After the training the final selection of temporary staff was made
on the basis of each applicants performance in a written test, which was given at the end of
the training. Regardless of the results of the test, no crossovers were allowed between re-
gions, except when it was deemed necessary by the office. Staff deployment in all regions
was done immediately after training. The first group in the field was that of regional and
team supervisors and listing clerks. The task was to list and to familiarize with the Primary
Sampling Units (PSUs) and do some publicity before interviewing. Many reasons war-
ranted office staff to do field monitoring of the data collection activities taking place in the
regions, including:
" The importance and uniqueness of the survey information for socio-economic plan-
ning for the country.
" Staff inexperience in conducting the budget survey.
" The temporary nature of the staff in the field.
- 45 -
" To retrain staff on aspects where mistakes had been identified.
" Respond to queries and attend to possible staff grievances.
To ensure that field operations went as smoothly as possible, field monitoring visits were
done at regular unannounced intervals. Monitoring teams spent on average two to three
days in each region before proceeding to the next region. Besides checking of question-
naires and general administration issues, monitoring teams re-interviewed some households
already covered by field staff and compared answers. Regional supervisors were required
to submit monthly reports about survey activities in their regions. A post-enumeration sur-
vey was conducted immediately after the main field work with the objective of testing the
values of information collected earlier. A refresher training of the best staff that partici-
pated in the main fieldwork was undertaken prior to the data collection of the PES
Table A-2: Distribution of sample households and sampling fractions by region and
urban/rural areas for the two surveys
1993/94 NHIES 2003/04 NHIES Region
Urban Rural Total Sampling
fraction
(%)
Urban Rural Total Sampling
fraction
(%)
Caprivi 48 240 288 1.6 300 480 780 4.6
Erongo 192 72 264 2.0 520 260 780 2.8
Hardap 144 96 240 1.8 300 480 780 5.1
Karas 168 72 240 2.0 400 380 780 5.0
Kavango 72 240 312 1.8 300 480 780 2.5
Khomas 648 48 696 2.1 1 040 260 1 300 2.2
Kunene 96 144 240 1.9 260 260 520 4.1
Ohangwena 0 432 432 1.5 260 520 780 2.2
Omaheke 72 168 240 2.5 260 260 520 4.1
Omusati 0 456 456 1.5 260 780 1 040 2.7
Oshana 168 264 432 2.0 400 640 1 040 3.5
Oshikoto 96 264 360 1.7 260 780 1 040 3.7
Otjozondjupa 240 168 408 1.9 360 420 780 3.1
Namibia 1944 2664 4608 1.8 4 920 6 000 10 920 3.1
Sampling
Stratified two-stage cluster sample design was used for the NHIES, where the first stage
units were geographical areas designated as PSUs and the second stage units were the
households. The first stage units were selected from the sampling frame of PSUs and the
second stage units were selected from a current list of households within each selected
- 46 -
PSU, which was compiled just before the interviews for the survey. The extensive stratifi-
cation of the frames together with the systematic sampling procedure enhanced the repre-
sentation of different types of sub-population groups in the NHIES sample.
PSUs were selected using probability proportional to size sampling coupled with the sys-
tematic sampling procedure where the size measure was the number of households within
the PSU in the 2001 Population and Housing Census. The households were selected from
the current list of households using systematic sampling procedure. The selected PSUs
were randomly allocated to the 13 survey rounds. The sample sizes were designed to
achieve reliable estimates at the national and regional levels. The distribution of the sample
households for the two surveys is given in Table A-2.
The number of households to be interviewed per PSU was fixed at 20. Increasing the num-
ber of sample households more than 20 in each PSU would not add much to the improve-
ment of the precision but would only increase costs. Population figures were estimated by
raising sample figures using sample weights. Sample weights were calculated based on
probabilities of selection at each stage. First stage weight was calculated using the sample
selection information from the sampling frame and the second stage weight was based on
sample selection information on the listing form. In the second stage, some households out
of the selected 20 households in a PSU did not participate in the survey due to refusals,
non-contact or non-completion of interview, etc. Such non-responding households were
few in number and there was no evidence to suggest that the excluded households were
significantly different from the responding ones. Hence, it was assumed that the non-
responding households were randomly distributed and the second stage weights were ad-
justed accordingly. The final sample weight was the product of the first and the second
stage weights, which were then incorporated into the database, so that inflating the sample
data would be automatically carried out when the tables were produced.
Changes in survey methods
In the 2003/2004 NHIES, the Central Bureau of Statistics took care not to depart unneces-
sarily from the methodology used in the previous survey in 1993/1994 to keep comparabil-
ity between the surveys to a maximum. Invariably however, surveys that are conducted 10
years apart will not be completely comparable. Methodologies do change over time and
improvements are introduced based on experiences and lessons learned. Moreover, a
household survey of this size, scope and complexity is a challenge for any statistics office
and even more so if faced with the severe capacity deficits that characterises the Central
Bureau of Statistics in Namibia. Nevertheless, the main differences between the two
NHIES conducted in 1993/1994 and in 2003/2004 were:
" The sample in 2003/04 comprised more than twice as many households. One impli-
cation is that sampling errors are reduced and estimates are thus statistically more
accurate. It follows that more lower level disaggregation of results can be done
without compromising robustness.
" In the most recent survey, a deliberate effort was made to improve the data collec-
tion especially when it came to reported consumption and income, and a larger
- 47 -
number of in-frequent annual expenditure items were collected directly in
2003/2004.
" To qualify as a household member in the 2003/2004 survey, a person would have
stayed in the household at least two weeks of a four weeks period. In the 1993/1994
survey, a person qualified as a household member having stayed at least one week
of a four weeks period.
" In the 2003/2004 survey, modern technology was used for data processing and data
cleaning, for instance the data on Form 1 was captured using digital scanning.
" The latest survey also included a module which measured the height and weight of
all household members in order to provide a basis for a comprehensive assessment
of the nutritional status of Namibians.
When comparing the results from the two surveys the effects of improved methodologies
and better coverage are difficult to separate from actual observed changes especially when
it comes to the income and expenditure data. Therefore, the Central Bureau of Statistics
generally advises that the users of the NHIES data treat observed changes over time be-
tween the two surveys as more indicative of direction rather than as precise estimates.
Work is currently underway to strengthen comparability between the two surveys and use
the 1993/1994 survey as a benchmark for further analysis of poverty and inequality in Na-
mibia.
- 48 -
ANNEX B: Poverty measures in Namibia and other
SADC countries
Poverty assessments typically begin by making two analytical choices. Firstly, a measure
of welfare or deprivation is selected, for instance income or consumption expenditure, and
secondly a thresholdpoverty lineis determined for that measure in order to distinguish
the poor from the non-poor. A poverty line can serve several useful purposes. It can be
used for monitoring poverty over time, tracking trends and changes in poverty levels; it can
be used for developing a poverty profile that describes the characteristics of the poor and
the environment in which they live; it can be used for targeting and defining entitlements
such as social grants; and it can be a focus for public debate around initiatives and policies
to fight poverty.
Poverty measures and poverty lines are generally either relative or absolute in nature. A
relative poverty line is determined from a cut-off point in the welfare distribution below
which a share of the population or households are located. Examples of such cut-offs vary
but are typically set at 30-50 percent, i.e. those with incomes of 30-50 percent below the
mean are considered poor. An absolute poverty line on the other hand is anchored explic-
itly in a specific level of welfare that is predetermined and which separates the poor from
the non-poor. The absolute poverty line is typically based on the Cost of Basic Needs
(CBN) required by households to meet a minimum daily nutritional requirement and cer-
tain essential non-food items (e.g. clothing and shelter). The main alternative method put
forward in the literature to the CBS approach is the Food Energy Intake (FEI) method,
which is not anchored in any bundle of goods but rather produces an estimate of the in-
come/expenditure level where the typical household is able to meet its nutritional require-
ments. However, a principal advantage of the CBN over the FEI approach is that it is wel-
fare consistent so that individuals with the same kind of living standards are treated equally
(Ravallion 1992; Ravallion and Bidani 1994).3
Both relative and absolute approaches have been applied in Namibia in the past. For in-
stance, Yaron et al (1992) used a 1989 food basket from Botswana to analyse poverty in
five communities in Namibias northern Okavango Region. The value of the basket was
adjusted for inflation and an adult equivalence scale was applied to account for differences
in the age and sex composition of households. On this basis, 40 percent of households were
found to be food insecure, i.e. not able to meet the costs of the food basket, and an addi-
tional 13 percent were found to not have sufficient income to cover the costs of the food
basket scaled up by one-third to allow for basic non-food needs. The sum, 53 percent, was
classified as generally poor. A similar approach was taken by Van Rooy et al (1994), in a
survey of 225 households in three communities representing different situations in Na-
mibia. The subsequent analysis tested the methodology for drawing a poverty line based on
the cost of basic needs approach (for simplicity the Botswana food basket was used again)
3 On the other hand, it has also been suggested that a single national food bundle may be inappropriate in set-
tings where the food consumption patterns of the poor are heterogeneous because of differences in the rela-
tive prices of staple foods (Tarp et al 2002).
- 49 -
and the standard Forster-Greer-Thorbecke (FGT) measures were calculated (more on these
measures below), and micro-simulations were done on the impact of cash transfers.
An alternative approach was followed in the analysis of the nation-wide 1999 Levels of
Living Survey (CBS 2001), which used a relative poverty line by defining the poor as all
those who had incomes of less than the national average. On that basis, 76 percent of
households in Namibia were classified as poor, with incidence in several regions above 90
percent. Unfortunately, this study did not elaborate on how this particular poverty line was
arrived at. In principle, a relative poverty measure is both simple and transparent, and can
be useful in identifying a population sub-group upon which to focus attention. However,
the approach has a couple of disadvantages (Lanjouw 2001). Firstly, the relative poverty
line is not particularly useful for some purposes such as measuring poverty over time and
space: irrespective of the income level, there will always be some in the population that
have incomes that are below the mean (except from the unlikely situation where everyone
has the exact same income, of course). If all incomes increase by the same proportion, the
poverty line will increase by the same proportion and the poverty measure will remain un-
changed. Similarly, the approach does not allow for comparisons across regions. Secondly,
the relative poverty line is essentially quite arbitrary and it is not clear from the 1999 sur-
vey why the poverty level was defined as the mean level of expenditure. Usually, applica-
tions of the relative poverty measure do not make use of some proportion of the mean, e.g.
50 per cent, which does not make it less arbitrary nor does it address the problems of com-
parability. However, in a high inequality society like Namibia average levels of income are
particularly unsuitable as an indicator of welfare. Also problematic, the study used un-
weighted data for the analysis, which meant that even if the survey sample was nationally
representative, the reported results were not.
International development agencies working with Namibia also weigh in on the application
of poverty measures in the country. For instance, the World Bank estimates poverty levels
using the share of the population that lives below daily poverty thresholds of US$1 ad-
justed for Purchasing Power Parities (PPP). The most recent estimate for Namibia is for
1993 with 35 percent of the population living on less than US$ 1 PPP per day (World Bank
2007). The US$ 1 poverty line is also the one that is used to monitor global progress to-
wards the first of the eight Millennium Development Goal to Eradicate Extreme Poverty
and specifically the target: Reduce by half the proportion of people living on less than a
dollar a day by 2015 compared to a 1990 base line. While the US$ 1 PPP poverty line
may be practical for international comparisons, it is less useful for national poverty meas-
urement especially since it relies on the conversion from national currencies into US$ using
PPP. So for country level monitoring, it is generally advised that countries use an official
threshold (or poverty line) set by the national government based on the specific characteris-
tics of the country (United Nations 2003). UN agencies traditionally include but go beyond
the money-metric approach and UNDP has defined a set of composite indices which in ad-
dition to income include educational and health outcomes to measure progress and setbacks
in human capabilities. These indices are published annually in the UNDP Human Devel-
opment Report using internationally comparable data for most of the worlds countries.
Moreover, the Human Development Index and Human Poverty Index (HPI) are calculated
for Namibia using official national data sources. The HPI for Namibia includes an income
- 50 -
poverty estimate, based on the national poverty line, in addition to the adult illiteracy rate
and the risk of dying before the age of 40 (Levine 2007).
All these efforts notwithstanding, as the custodian of the statistics system and ultimately
responsible for setting national standards in social and economic statistics, the official pov-
erty line for Namibia is determined by the Central Bureau of Statistics. In the two previous
NHIES reports, poverty was defined using a food-share approach, which is a variant of the
absolute poverty measures. The food-share approach is based on the empirical observation
of an inverse relationship between overall household incomes and the share spent on food,
which implies that relatively poorer households spend a higher proportion of their total
consumption expenditure on food compared to more well-off households.4 On that basis,
the Central Bureau of Statistics defined poor households as those spending 60 percent or
more of total consumption expenditure on food, and the severely poor as those spending
80 percent or more. These cut-off points have been used in both NHIES and have served as
the official poverty lines referred to in major strategies and policies for national develop-
ment and poverty reduction. The origin of this specific poverty measure as well as the
methodological justification for choosing it during the first NHIES has not been docu-
mented.
Figure B-1: Non-parametric Engel curve for Namibia
40
45
50
55
0 100 200 300 400 500 600 700
Adult equivalent monthly expenditure (N$)
F
o
o
d
s
h
a
re
Figure B-1 shows the relationship between the food-share of household expenditure and the
monthly adult equivalent expenditure from the NHIES. Two features stand out from the
graph. Firstly, that the data appears to confirm the inverse relationship between total con-
sumption and food consumption. Secondly, that this does not hold at the lowest consump-
4 This relationship is often referred to as Engels Law after the 19th Century Prussian statistician Ernst En-
gel who in a study of the budgets of Belgian worker families concluded a.o.: The poorer is a family, the
greater is the proportion of the total outgo which must be used for food (see Zimmerman 1932).
- 51 -
tion levels where the food-share appears to be rising with rising income. This lends support
to the unity elasticity observation and implies that while there may be a general relationship
between level of income and the share spent on food, using a cut-off point in the food-share
distribution to identify the poor and the poorest of the poor is problematic.
On Figure B-2, the distribution of total monthly and adult equivalent expenditure of house-
holds are illustrated for those households with a food-share of 60 percent or more. The figure
shows how the food-share method identifies households as poor even if they have adult
equivalent expenditure levels of more than N$ 1,000 and up to as high as N$ 7,000. The av-
erage for the group is N$ 382. Figure B-3 makes the same comparison but for households
with food-shares of 80 percent or more. This classification used to identify severely poor
includes households with adult equivalent expenditure as high as N$ 3,800, and the average
expenditure for this group is even higher (N$ 411) than for those with 60 percent food-share.
While setting poverty lines invariably involves a degree of arbitrariness, there appears to be a
problematic misspecification of households using the food-share method and this misspeci-
fication is particularly problematic when it comes to the poorest of the poor.
Figure B-2: Expenditure distribution among households with food-share 60% or
more
Per capita monthly expenditure
7,000
6,500
6,000
5,500
5,000
4,500
4,000
3,500
3,000
2,500
2,000
1,500
1,000
500
0
Cases weighted by SWEIGHT
N
u
m
b
e
r
o
f
h
o
u
s
e
h
o
ld
s
2,000
1,000
0
Std. Dev = 342.85
Mean = 338
N = 2619.72
The empirical literature on poverty measurement covers a range of strengths and weak-
nesses related to the food-share method. For instance, Ravallion (1992) notes that a major
drawback of setting a poverty line using the food-ratio method is that the relationship be-
tween the food-share and consumption will generally differ across households for reasons
unrelated to poverty rather reflecting differences in the relative prices, tastes and availabil-
ity. Moreover, the income elasticity of demand for food can be close to unity for very poor
- 52 -
households, which renders the indicator unreliable. However, food-share data can some-
times provide a useful supplementary test, particularly if one is worried about the quality
of, for example survey data or the price deflator. Ravallion and Huppi (1991) thus find that
in applying the food-share data, the same qualitative conclusions are arrived at in compar-
ing poverty over time and across sectors in Indonesia as the ones reached for consumption
and income data. This is taken as adding strength to the conclusion of the paper that pov-
erty in Indonesia had declined. Nevertheless, as noted by Deaton (1997:2): even if our
main concern is with food, and if we believe that food consumption is a rough but useful
measure of welfare, why focus on the share of food in the budget in preference to more di-
rect measures such as food consumption or nutrient intake?
Figure B-3: Expenditure distribution among households with food-share 80% or
more
Per capita monthly expenditure
3800
3600
3400
3200
3000
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
600
400
200
0
Cases weighted by SWEIGHT
N
u
m
b
e
r
o
f
h
o
u
s
e
h
o
ld
s
200
100
0
Std. Dev = 416.76
Mean = 411
N = 322.49
While moving from the food share method to the CBN approach in Namibia would make
methodological sense, it would also bring the country closer to the methodologies applied
in other countries. United Nations Statistics Division (2005) finds that the large majority of
developing countries follow the CBN approach in producing income or expenditure based
poverty statistics. Looking specifically at the SADC region, it is clear that Namibia stands
out in its present choice of poverty measure. Table B-1 provides a comparison between
Namibia and a selection of countries in the sub-region. Several features are noteworthy.
Most significantly, it is clear how the majority of countries apply an absolute measure in
the form of the CBN approach to setting the official poverty line. Major exceptions are
Mauritius, which uses a relative measure as is often the case for more developed societies
and Namibia with its food-share method. It is important to also to note that while most
countries apply a CBN approach, there are great variations in its application, notably how
- 53 -
the food basket is determined, how adjustments are made for adult equivalents and how the
non-food components are determined. On this basis, it can be concluded that by adopting a
CBN approach to poverty measurement, Namibia would be more aligned with international
practices, and that even then, the country could enjoy considerable methodological flexibil-
ity in setting its poverty line.
Table B-1: Definitions of income poverty in selected SADC countries
Country Official poverty line definition Poverty incidence Latest data source (previ-
ous surveys)
Botswana CBN; National Poverty Datum Line
and US$ 1 PPP
Poverty Datum Line:
30.3%; US$ 1: 23.4% (p)
2002/03 Household Income
and Expenditure Survey
(1993/94)
Lesotho CBN: food basket adjusted for adult
equivalent, and non-food
Very poor (food poor):
29.1%; Poor: 50.2% (hh)
2002/03 Household Budget
Survey (1994/95)
Tanzania CBN: food basket (RDA 2200 kcal)
and non-food; consumption adjusted
for adult equivalent.
Food poverty: 18.7%; Ba-
sic needs poverty: 35.7%
(p)
Household Budget Survey
2000/01 (1991/92)
South Africa* CBN: food basket (RDA 2261 kcal)
and non-food
Lower bound: 52.6%; Up-
per bound: 70.4%
2000 Income and Expendi-
ture Survey (1995)
Swaziland CBN: food basket (RDA 2100 kcal)
and non-food; consumption adjusted
for adult equivalent.
Extreme (food) poverty:
33%; Poverty: 69%
2000-2001 Swaziland
Household Income and Ex-
penditure Survey (1994/95)
Malawi CBN: food basket and non-food Ultra (food) poor: 22.3%;
Poor: 52.4% (p)
Second Integrated House-
hold Survey 2005 (1998)
Mauritius Relative poverty line set at the half the
median household income per adult
equivalent
Poor: 8% (hh) Household Budget Survey
2006/07 (2001/02, 1996/97)
Namibia Food expenditure as share of total ex-
penditure
Poor (food-share is 60% or
more): 27.9%; Severely
poor (80% or more): 3.9%
(hh)
2003/2004 Namibia House-
hold Income and Expenditure
Survey (1993/94)
Zambia CBN: food basket and non-food; con-
sumption adjusted for adult equivalent.
Extreme poverty: 57.9%;
Overall poverty: 72.9%
Living Condition Monitor-
ing Survey 1998 (1996, 1993,
1991)
Zimbabwe CBN: food basket and non-food Population below food
poverty line (Extreme pov-
erty): 58%; Population
below poverty line
(Poor+Very poor); 72% (p)
2003 Poverty Assessment
Study Survey (1995)
* Not yet formally adopted by the statistics office.
Note: CBN=cost of basic needs approach, (p) = persons, (hh) = households,
Sources: Information released in official printed or online survey reports or national poverty reduction
strategies.
- 54 -
ANNEX C: Setting the poverty line
Setting up an absolute poverty line for Namibia using the CBN approach has been a fairly
labour intensive process and has included a series of methodological steps. Each of these
steps have been made after careful analysis and review of the extensive international litera-
ture, extensive discussions among the technical team members and outside experts, and de-
cisions ultimately sanctioned by the management of the Central Bureau of Statistics. In the
following, the process of developing a new poverty line for Namibia is broken down into a
series of methodological steps that follow once the decision has been made to adopt a CBN
approach. This work precedes that of Van Rooy et al (2006), which first applied the meth-
odology to the 1993/1994 survey.
Step 1: Determine the energy requirements
The first step in setting the poverty line is to determine the cut-off point or threshold of ba-
sic needs. A poverty line that uses a CBN approach typically emphasises food as the most
basic need and it is therefore linked to a minimum level of calorific requirements. These
requirements should ideally be determined through national nutritional studies and be com-
puted for different groups of persons defined according to sex, age and level of activity (see
FAO/WHO/UNU 1985). One review of country experiences in setting energy thresholds
found a range in the applications from 2,000 kcal in the Maldives to 3,000 kcal in Uganda
(UNSD 2005). For the purposes of this report, the minimum calorific requirement is set at
2,100 kcal on average per person. This is the standard currently used by the Ministry of
Health and Social Services (MoHSS) and international agencies, for instance the World
Food Programme (WFP) uses this standard/measurement when determining emergency
food aid. It is also the value that is used by several other middle income countries including
Thailand, Turkey and Swaziland. It should be noted that sensitivity tests showed that vary-
ing the calorific threshold from 2000-2300 kcal had limited impact on poverty levels, sug-
gesting that the poverty line is quite robust to the choice of calorific threshold. Neverthe-
less, it is recommended that using the extensive nutrition data as collected during the
NHIES, a new and more detailed scale of energy requirements be developed. This is one
among several points for follow-up identified as part of this analytical process.
Step 2: Select the reference group
The second step towards a CBN-based poverty line is to select the reference group. This
basically involves choosing between households and individuals, and if the former is cho-
sen, adjusting for differences in age composition. Adult equivalence scales recognise that,
for example, a household composed of four adults need a different level of resources than a
household composed of one adult and three children in order to reach the same level of
economic welfare. On an aggregate level, the use of equivalence also helps adjust for
changes over time in the structure of the population, which is particularly relevant in Na-
mibias case of falling fertility since 1990. In the NHIES, children under the age of 5 are
assigned a weight of 0.5 in terms of adult equivalent needs and children between 6 and 15
are assigned a weight of 0.75. Adults 16 years and over, irrespective of gender are assigned
- 55 -
a weight of 1.5 This scale was used in the poverty analysis presented in this report.6 The
choice of equivalence scales and economies of scale parameters can make quite a differ-
ence and during the course of the analysis several scales were tested. In the analysis pre-
sented in this report, no adjustment is made for economies of scale but this issue could be
revisited in the future. Again, the nutrition data of the NHIES could be used to develop a
more detailed set of equivalence scales.
Step 3: Determine the contents of the food basket
The next step involves selecting the specific goods that should go into the food basket.
There are a number of ways of doing this. In the present analysis, the food basket is based
on the top purchased items of households in 2nd to 4th consumption expenditure deciles in
the survey. The bottom decile of the distribution (i.e. the 10 percent with the lowest expen-
diture levels) were excluded to eliminate outliers. On this basis, the top 30 purchased
food/beverage items were selected for the food basket together with the 15 most commonly
consumed in-kind food items (e.g. from own-production). Sensitivity analysis showed that
the results were robust to changes in the specification of the poverty basket to allow for
more items. This approach for selecting the food basket was preferred because it is based
on the actual consumption patterns of the lower deciles of expenditure distribution. This
way it is ensured that the food and beverage items in the basket are consistent with local
tastes and preferences. Moreover, very expensive, luxury-type food items, unlikely to be
consumed by the poor are not heavily represented in the basket. The specific items in the
national food basket are found in Annex D.7
Step 4: Set price of food items
The prices for each of the items in the food basket was determined by using information
from the collection of the Consumer Price Index (CPI). In the few cases where the CPI
does not include the specific item prices, the daily record books were used. Regional price
differences were accounted for in the compilation of the food basket by using CPI data.
5 The source of this scale is still unknown. The NHIES 1993/94 Report makes reference to: SSD Research
Report 10 (1994), UNAM, February. Another study suggests that the scale in use emanates from the Bot-
swana poverty datum line. See: Ekström, E (1998), Income Distribution and Labour Market Discrimination:
A Case Study of Namibia, Research Institute of Industrial Economics (IUI), IUI Working Paper Series No.
502, October. Here a reference is also made to: Central Statistics Office, The distribution of economic re-
sources in the population of Namibia, Some highlights, 1995, National Planning Commission, Windhoek.
6 Often in studies of poverty an additional adjustment is made for economies of scale arising as the size of the
household increases however there exists little guidance for choosing the value of the parameter (White and
Masset 2003; Deaton and Paxson 1998; Lanjouw and Ravallion, 1995).
7 The team also conducted some preliminary experimental analysis on more region specific food baskets. This
work revealed some diversity in the food consumption habits of the poor. For regions such as Ohangwena,
Oshikoto and Oshana only a few items were not the same as in the national basket but for less populous re-
gions of Hardap and Karas in the south of the country more than a third of the purchased items and more than
two thirds of in-kind items changed. This will affect the food poverty line especially for these regions. More-
over, region specific differences in non food expenditure appear important in especially pushing up the upper
bound poverty line in urban regions such as Khomas and Erongo. As a follow-up to this report on the national
poverty line the team will pursue additional analysis that will aim to investigate these regional differences
further. It will be critical to account for differences in the level of the region specific poverty lines that are
attributable to preferences and tastes, availability of certain foods, their relative prices, as well as the standard
of living among the poor.
- 56 -
Since 2004, collection points for the CPI have been expanded beyond Windhoek to include
8 regional groupings although data is still only collected in urban areas. There is an impor-
tant lesson for the next NHIES to include a survey of prices.
Step 5: Calculate the food poverty line
With all this information at hand, the food poverty line can be calculated. First, the average
expenditure per capita for each of the 45 items are converted into daily calorific values us-
ing nutritional data on calorie content per gram. Then the costs for each household in meet-
ing the daily calorific minimum of 2,100 kcal for its members can be calculated. This
represents the food poverty line and based on the 2003/2004 NHIES, this was calculated as
N$ 127.15
Step 6: Include non-food items
While having sufficient resources in the household to meet food requirements is critical, it
is not enough for the poverty classification. This is so because households that can afford to
meet food requirements of all members, but lack resources to purchase clothing and shelter,
for example, should likely be considered deprived in a very basic sense. There are several
ways of including these essential non-food items. Two approaches stand out. Under the
first approach, non-food expenditure is calculated from actual expenditure on non-food
items by households with food expenditure approximately equal to the food poverty line.
Under the second approach, non-food expenditure is calculated from actual non-food ex-
penditure of households whose consumption expenditures are equal to the food poverty
line. The rationale for the latter, more austere approach is that if these households have the
ability to obtain the minimum food basket, but choose to divert resources to buy non-food
items, then the household must clearly view these items as essential. In the literature, both
methods are found to be methodologically sound and they are often considered together as
a lower and upper bound, respectively (Ravallion 1998). In the subsequent poverty analysis
for Namibia, both measures are applied and should be interpreted as representing a range
of poverty in the country. Since no group of people have total expenditure, or food expen-
diture, exactly equal to the food poverty line, a simple nonparametric procedure was used.
The median non-food expenditure per capita was calculated for households with per capita
total expenditure in a small interval (plus or minus one percent) around the food poverty
line. Successively, larger intervals were selected, a total of five times so that the largest in-
terval is ±5%, and a simple average was taken of the five observations of median non-food
expenditure per capita around the food poverty line.8 Table C-1 shows the values of the
food poverty line as well as the upper and lower poverty lines for the 2003/2004 NHIES.
Table C-1: Annual values of poverty lines, monthly N$ per capita
Poverty line 2003/2004
Food poverty line 127.15
Lower bound poverty line: severely poor 184.56
Upper bound poverty line: poor 262.45
8 This approach was proposed by Ravallion (1998) and applied in e.g. Nepal (Lanjouw 2001) and Lesotho
(May and Roberts 2005).
- 57 -
Step 7: Choose measures for analysis
Once the poverty lines have been determined, the final step is to select the measures to ex-
press the shortfall and deprivation. The first poverty measure to define is the poverty head-
count or incidence of poverty. This is the share of the population that has an income y that
is less than the poverty line z.
If the population size is n and the share of poor people is q, then the poverty headcount is
given by:
H =
n
q
However, as a poverty measure H has some limitations because it does not recognise the
size of the aggregate income shortfall of the poor as well as the distribution of income
among the poor. As has become standard in poverty research, the analysis presented for
Namibia uses the more general Foster-Greer-Thorbecke (FGT) class of poverty measures
given by:
= ú
ú
û
ù
ï
ï
ð
î
=
q
i
i
z
z
n
P
y
1
1
±
±
While an infinity of poverty measures can be derived depending on the value of the pa-
rameter ±, three measures are of particular interest:
1. In the case that ± equals 0, then we have P0 = H , i.e. the poverty headcount
measure.
2. P1 (± equals 1) is referred to as the poverty gap measure and indicates the aver-
age aggregate consumption expenditure shortfall, or depth of poverty, of those
below the poverty line.
3. P2 (± equals 2) is the squared poverty gap and referred to as the severity of pov-
erty as it places greater weight on those that are further from the poverty line.
The Foster-Greer-Thorbecke set of indices has the agreeable feature that the indices may be
decomposed. This way, one may calculate how large a share of the contribution to poverty
a subgroup of the population make.
- 58 -
ANNEX D: The national food basket
Table D- 1: Purchased items
Ave
annual
exp (N$)
Ave
monthly
exp (N$)
No. HH
consuming
item
No. HH not
consuming
item
% con-
suming
item
N*Ave exp
1. Maize meal/grain/samp 641.00 53.42 64434 47074 58 3441823
2. Beef 327.59 27.30 70643 40865 63 1928498
3. Sugar, all types 270.89 22.57 72151 39357 65 1628715
4. Bread (all types) 107.26 8.94 67698 43809 61 605129
5. Frozen fish 145.80 12.15 42827 68681 38 520347
6. Cooking oil 102.50 8.54 53251 58257 48 454835
7. Rice 85.60 7.13 38808 72700 35 276833
8. Soft drinks 73.96 6.16 38184 73323 34 235353
9. Fresh fish 66.27 5.52 27715 83793 25 153062
10. Mahangu meal/grain/samp 93.55 7.80 17618 93890 16 137345
11. Powdered soup 37.89 3.16 38828 72680 35 122591
12. Chicken 68.52 5.71 20104 91404 18 114800
13. Local home-made brew, all
types (Ombike, tombo, ka-
shipembe)
39.63 3.30 31912 79596 29 105400
14. Beer/ales/ciders 62.22 5.19 17218 94290 15 89276
15. Breads, cake flour (all types) 67.34 5.61 15514 95994 14 87059
16. Macaroni, spaghetti, noodles 42.65 3.55 24355 87152 22 86558
17. Fresh milk 42.64 3.55 24150 87358 22 85810
18. Potatoes, English 45.97 3.83 19352 92156 17 74126
19. Sweets 23.36 1.95 32892 78615 29 64023
20. Tea 30.28 2.52 23779 87729 21 59993
21. Bottled/Tinned fish 34.38 2.86 20901 90607 19 59874
22. Goat meat 41.53 3.46 16589 94919 15 57413
23. Traditional sour milk 22.07 1.84 15603 95905 14 28702
24. ONION 14.02 1.17 23943 87565 21 27982
25. Tomatoes 16.34 1.36 20168 91340 18 27463
26. Salt 11.95 1.00 24704 86804 22 24597
27. Coffee 17.47 1.46 16503 95004 15 24022
28. Dried fish 19.07 1.59 13631 97877 12 21667
29. Vetkoek 11.12 0.93 23277 88231 21 21564
30. Fruit juice and squashes 16.79 1.40 15236 96272 14 21315
- 59 -
Table D-2: In kind items
Ave an-
nual exp
(N$)
Ave
monthly
exp (N$)
No. HH con-
suming item
No. HH not
consuming
item
% consum-
ing item
N*Avexp
31. Mahangu meal/grain/samp 1160.37 96.70 67218 44290 60 6499846
32. Spinach/ombindi/derere/
mutete/ekaka
186.61 15.55 66867 44641 60 1039822
33. Maize meal/grain/samp 232.46 19.37 33340 78167 30 645865
34. Beef 163.88 13.66 38445 73063 34 525027
35. Beans (dried) 106.81 8.90 34348 77160 31 305713
36. Chicken 108.96 9.08 30263 81245 27 274783
37. Magau/Oshikundu 122.96 10.25 24648 86860 22 252551
38. Goat meat 146.91 12.24 19750 91757 18 241802
39. Traditional sour milk 68.14 5.68 24315 87193 22 138066
40. Baby marrows (squash)
Pumpkins and squashes, all
types
75.33 6.28 20007 91501 18 125596
41. Beans (fresh) 107.31 8.94 13763 97745 12 123074
42. Fresh milk 73.50 6.13 18952 92555 17 116086
43. Local home-made brew, all
types (ombike, tombo, ka-
shipembe)
56.55 4.71 20435 91073 18 96296
44. Ground nuts/Eefukwa 61.18 5.10 13218 98290 12 67392
45. Fresh fish 44.40 3.70 12956 98552 12 47938
- 60 -
ANNEX E: First-order stochastic dominance tests
Setting the poverty line invariably involves an element of arbitrariness as to where the cut-
off that separates the poor from the non-poor is eventually made. In Annex C, it was de-
scribed how during the design of the poverty line, various tests for robustness were con-
ducted. In this Annex, an important additional test for robustness of the poverty measure is
conducted using graphical techniques and the theory of stochastic dominance. By plotting
the cumulative function of household expenditure, sometimes called the poverty incidence
curve, of different subgroups of households, it is possible to assess whether the ranking of
these groups in terms of poverty levels are robust with respect to the poverty line.
An example is given on Figure E-1, which plots the cumulative distribution functions of
urban and rural households in Namibia. For a given level of household expenditure on the
horizontal axis, reading off the vertical axis for one of the curves indicates the incidence of
poverty, which would result if a poverty line equal to that expenditure level had been se-
lected. For example, the upper-bound poverty line of N$ 262.45 implies a headcount rate of
almost 40 percent in rural areas and just over 10 percent in urban areas. What is more, at
any given level of the poverty line, the poverty headcount will be higher in the rural areas
compared to urban areas. Since the curve representing urban at all points lies below the
curve representing rural without any point of intersection, then following Foster and Shor-
rocks (1988), it can be stated that the former dominates the latter in the first order and it
can be concluded that poverty as measured by any of the FGT measures in rural Namibia is
higher than in urban Namibia, irrespective of where the poverty line is drawn. A similar
conclusion can be drawn when it comes to other background variables. On Figure E-2, the
cumulative distribution functions of male- and female-headed households are plotted. The
difference in poverty levels between the two sexes is evidently small, but nevertheless, the
conclusion that female-headed households are poorer is visibly robust to the specification
of the poverty line. Figure E-3 and Figure E-4 provide further illustrations of first-order
stochastic dominance, confirming that the conclusions regarding the linkages between the
poverty status of the household and the level of education and main source of income are
robust to the of the specification of the poverty line.
The type of stochastic dominance described here seizes to exist if cumulative distribution
curves intersect at some point. Then, it is no longer the case that the same ranking of pov-
erty would remain over all possible poverty lines and FGT measures. This is the case when
it comes to the regions in Namibia as exemplified in Figure E-5. Each line represents the
cumulative distribution function for one of the 13 administrative regions. The performance
of three regions is highlighted for illustrative purposes. Firstly, it is clear that Erongo
dominates all other regions, irrespective of where the poverty line is set. Secondly, at
higher levels of expenditure, there are a number of points of intersection and this affects the
ranking. For instance, if the poverty line is set below N$ 500, then Caprivi is ranked 5th in
terms of poverty headcount but thereafter the rank rises above several regions, including
Oshikoto as indicated, and at a poverty line above N$ 1,200 Caprivi is ranked 1st.
- 61 -
Figure E-1: Cumulative distribution functions for urban and rural households
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500 2000 2500 3000
Household expenditure (in N$ and monthly adult equivalents)
P
o
o
r
a
s
p
ro
p
o
rt
io
n
o
f
to
ta
l
p
o
p
u
la
ti
o
n
Urban
Rural
Upper bound
poverty line
Figure E-2: Cumulative distribution functions for male and female-headed house-
holds
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500 2000 2500 3000
Household expenditure (in N$ and monthly adult equivalents)
P
o
o
r
a
s
p
ro
p
o
rt
io
n
o
f
to
ta
l
p
o
p
u
la
ti
o
n
Male
Female
Upper bound
poverty line
- 62 -
Figure E-3: Cumulative distribution functions by highest level of education attained
by the head of household
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500 2000 2500 3000
Household expenditure (in N$ and monthly adult equivalents)
P
o
o
r
a
s
p
ro
p
o
rt
io
n
o
f
to
ta
l
p
o
p
u
la
ti
o
n
Upper bound
poverty line
Tertiary
Secondary
Primary
No formal
Figure E-4: Cumulative distribution functions by main source of income
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500 2000 2500 3000
Household expenditure (in N$ and monthly adult equivalents)
P
o
o
r
a
s
p
ro
p
o
rt
io
n
o
f
to
ta
l
p
o
p
u
la
ti
o
n
Upper bound
poverty line
Commercial farming
Salaries and wages
Household business
Pensions
Subsistence
agriculture
- 63 -
Figure E-5: Cumulative distribution functions by region
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500 2000 2500 3000
Household expenditure (in N$ and monthly adult equivalents)
P
o
o
r
a
s
p
ro
p
o
rt
io
n
o
f
to
ta
l
p
o
p
u
la
ti
o
n
Upper bound
poverty line
Khomas
Caprivi
Oshikoto
- 64 -
ANNEX F: Poverty profile tables
Table F-1: Incidence, depth and severity of poverty by region and urban/rural (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Caprivi 12.5 3.4 1.4 4.5 28.6 8.7 3.7 5.2 18607
Erongo 4.8 1.5 0.6 2.6 10.3 3.3 1.5 2.8 27713
Hardap 21.9 7.3 3.3 7.0 32.1 13.1 6.9 5.1 16365
Karas 12.5 4.2 2.0 3.8 21.9 8.1 4.1 3.3 15570
Kavango 36.7 12.5 6.0 23.2 56.5 23.0 12.1 17.8 32354
Khomas 2.4 0.5 0.2 3.0 6.3 1.6 0.6 4.0 64918
Kunene 13.1 4.1 1.9 3.4 23.0 8.3 4.1 3.0 13365
Ohangwena 19.3 4.4 1.4 14.2 44.7 12.7 5.0 16.5 37854
Omaheke 17.5 5.5 2.6 4.5 30.1 10.8 5.4 3.9 13347
Omusati 12.8 3.3 1.3 9.8 31.0 8.5 3.6 11.9 39248
Oshana 7.8 1.7 0.6 4.8 19.6 5.2 2.0 6.1 31759
Oshikoto 16.6 3.7 1.4 10.3 40.8 11.1 4.4 12.7 31871
Otjozondjupa 15.8 4.9 2.1 8.8 27.8 9.9 4.8 7.8 28707
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Urban 6.0 1.8 0.8 17.7 12.0 3.9 1.9 17.7 150532
Rural 19.1 5.4 2.3 82.3 38.2 12.3 5.6 82.3 221145
- 65 -
Table F-2: Incidence, depth and severity of poverty by sex and age of household head (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Sex of head of household
Female 15.1 4.4 1.9 44.4 30.4 9.9 4.6 44.6 150451
Male 12.9 3.6 1.5 55.3 25.8 8.3 3.8 55.2 219709
Not specified 12.4 2.1 0.5 0.4 15.0 5.6 2.3 0.2 1518
Age of head of household
16-20 14.4 4.5 1.8 1.7 22.5 8.8 4.3 1.3 6041
21-24 10.8 2.6 1.0 3.2 19.1 6.3 2.8 2.9 15349
25-29 8.3 2.3 1.0 5.9 18.2 5.4 2.5 6.4 36081
30-34 7.5 2.5 1.1 6.9 17.9 5.5 2.6 8.2 46835
35-39 10.0 2.8 1.2 9.4 18.7 6.2 2.8 8.7 47878
40-44 12.4 3.7 1.7 10.5 23.1 8.0 3.8 9.8 43390
45-49 12.1 3.4 1.5 8.0 22.4 7.6 3.5 7.4 34040
50-54 12.1 3.6 1.5 7.2 27.0 8.2 3.7 8.1 30795
55-59 18.3 5.7 2.7 7.9 35.4 12.1 5.8 7.6 22158
60-64 23.6 6.5 2.7 10.7 42.6 14.7 6.8 9.6 23194
65+ 22.7 6.1 2.5 28.1 47.5 14.6 6.5 29.5 63629
Don't know 12.9 3.5 1.4 0.6 20.9 7.0 3.4 0.5 2288
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-3: Incidence, depth and severity of poverty by main language spoken in household
(%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
San 39.0 14.2 7.0 3.8 59.7 24.9 13.5 2.9 4967
Caprivi languages 10.8 2.8 1.1 4.1 24.6 7.4 3.1 4.7 19664
Otjiherero 8.8 2.5 1.1 5.6 17.0 5.6 2.6 5.4 32686
Rukavango 34.9 11.7 5.5 23.6 54.4 21.8 11.4 18.4 34748
Nama/Damara 21.4 7.4 3.4 17.7 34.2 13.6 7.1 14.2 42484
Oshiwambo 11.8 2.7 1.0 41.6 28.5 7.8 3.1 50.5 181395
Setswana 1.0 0.1 0.0 0.0 14.5 2.2 0.5 0.2 1479
Afrikaans 3.5 1.1 0.5 2.7 7.9 2.3 1.1 3.0 39374
German .. .. .. .. .. .. .. .. 4005
English 0.4 0.1 0.0 0.1 0.6 0.2 0.1 0.0 6889
Others 9.6 2.1 0.8 0.7 16.4 5.4 2.3 0.6 3984
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 66 -
Table F-4: Incidence, depth and severity of poverty by main source of household income (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Salaries/Wages 6.6 1.9 0.8 22.3 13.8 4.3 2.0 23.1 172254
Subsistence
Farming
17.6 4.4 1.8 36.9 40.3 11.7 4.9 42.3 107519
Commercial
Farming
.. 0.0 0.0 .. 2.6 0.6 0.1 0.1 2753
Non-Farming
Business
13.7 3.9 1.6 6.6 24.1 8.4 3.9 5.8 24802
Pensions 28.4 8.4 3.6 18.9 49.6 17.7 8.5 16.5 34159
Cash Remittances 23.1 7.5 3.5 3.8 35.5 13.8 7.2 2.9 8468
Rental Income .. .. .. .. .. .. .. .. 819
Interest from
Savings/ Invest-
ments
8.9 2.2 0.6 0.1 8.9 4.2 2.0 0.1 633
Maintenance
grants
23.6 8.2 3.7 0.9 38.5 15.2 7.8 0.8 2049
Drought relief
assistance
53.6 19.0 9.3 1.5 66.0 31.8 17.8 0.9 1423
In kind receipts 25.8 9.4 4.8 3.7 41.1 17.2 9.3 3.0 7391
Other 34.1 11.2 4.9 4.1 56.6 21.2 10.8 3.4 6123
No Income 32.5 12.0 5.9 0.6 57.6 22.8 11.8 0.5 890
Not stated 14.0 5.3 2.5 0.7 28.0 10.9 5.4 0.7 2396
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-5: Incidence, depth and severity of poverty by main source of household income (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Salaries/wages 6.6 1.9 0.8 22.4 13.8 4.3 2.0 23.3 172254
Subsistence
farming
17.6 4.4 1.8 37.1 40.3 11.7 4.9 42.6 107519
Commercial
farming
0.0 0.0 .. 2.6 0.6 0.1 0.1 2753
Non-farming
business
13.7 3.9 1.6 6.7 24.1 8.4 3.9 5.9 24802
Pensions 28.4 8.4 3.6 19.0 49.6 17.7 8.5 16.6 34159
Cash remittances 23.1 7.5 3.5 3.8 35.5 13.8 7.2 2.9 8468
In-kind receipts 25.8 9.4 4.8 3.7 41.1 17.2 9.3 3.0 7391
Other 30.7 10.3 4.6 6.7 47.5 18.9 9.8 5.2 11047
No income 32.5 12.0 5.9 0.6 57.6 22.8 11.8 0.5 890
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 369282
Note: The categories in the former table were collapsed to produce the second due to the small sample sizes in some
of the cells. For instance, those reporting drought relief assistance as their main income source only number 40
households (unweighted n).
- 67 -
Table F-6: Incidence, depth and severity of poverty by type of dwelling unit (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Detached House 3.5 0.9 0.3 8.6 8.1 2.3 1.0 10.0 126368
Apartment 1.3 0.4 0.2 0.3 2.7 0.9 0.4 0.3 11792
Traditional Dwelling 21.3 5.8 2.5 67.5 43.6 13.8 6.2 69.2 162784
Improvised Housing 18.5 5.9 2.7 22.2 31.9 11.7 5.8 19.2 61716
Other 8.2 2.7 1.2 1.4 14.8 5.0 2.6 1.3 9017
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-7: Incidence, depth and severity of poverty by type of tenure (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owned 18.8 5.4 2.3 88.3 36.9 12.1 5.6 86.8 241125
Owned but not paid off 1.9 0.6 0.2 1.6 4.2 1.3 0.6 1.7 42628
Occupied free 9.5 2.6 1.1 7.8 21.8 6.4 2.8 8.9 41913
Rented w/o subsidy 2.9 0.8 0.3 2.2 6.5 2.0 0.8 2.5 39126
Rented with subsidy 1.0 0.1 0.0 0.1 1.4 0.4 0.1 0.1 6791
Other .. .. .. .. .. .. .. .. 54
Not specified .. .. .. .. .. .. .. .. 42
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-8: Incidence, depth and severity of poverty by material for roof (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Cement blocks 1.8 0.2 0.0 0.1 6.3 1.1 0.3 0.2 2438
Bricks 7.4 2.2 0.7 0.2 22.5 5.9 2.3 0.2 1049
Iron/Zinc 9.3 2.8 1.2 37.1 18.0 6.0 2.8 35.8 203568
Poles/sticks/grass 19.2 4.9 2.0 9.7 40.3 12.4 5.4 10.2 25971
Sticks/mud/clay/dung 18.1 4.8 1.9 0.8 29.0 10.5 5.0 0.6 2254
Asbestos 1.3 0.4 0.2 0.4 5.1 1.2 0.5 0.9 17240
Tiles .. .. .. .. .. .. .. .. 928
Slate 13.0 2.3 0.6 0.2 17.1 6.2 2.5 0.1 815
Thatch 23.2 6.5 2.8 50.2 47.0 15.0 6.8 50.9 110990
Other 11.0 3.3 1.6 1.3 17.9 6.7 3.4 1.1 6165
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 68 -
Table F-9: Incidence, depth and severity of poverty by material for the wall (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Cement blocks 3.7 0.9 0.3 10.4 9.4 2.6 1.0 13.3 145317
Bricks 7.2 1.4 0.4 1.4 16.6 4.4 1.6 1.6 9905
Iron/Zinc 19.8 6.4 2.9 22.0 33.3 12.4 6.2 18.6 57029
Poles/sticks/grass 18.9 4.8 2.0 20.4 41.3 12.1 5.2 22.3 55328
Sticks/mud/clay/dung 23.1 6.4 2.7 38.9 46.1 14.9 6.7 38.8 86236
Asbestos 5.4 2.2 1.2 0.2 19.5 4.8 2.3 0.4 2316
Tiles 12.7 2.9 1.0 0.2 12.7 5.8 2.8 0.1 908
Slate .. 0.0 0.0 .. 17.2 2.7 0.4 0.0 186
Thatch 36.1 13.7 6.8 3.5 49.2 22.4 12.7 2.4 4912
Other 16.1 6.1 3.3 2.9 27.0 10.7 6.0 2.4 9077
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-10: Incidence, depth and severity of poverty by material for the floor (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Sand 21.8 6.5 2.9 57.0 41.9 14.0 6.6 54.7 133987
Concrete 5.1 1.3 0.5 17.6 11.3 3.3 1.4 19.5 177125
Mud/clay/dung 22.6 6.2 2.5 24.8 46.1 14.7 6.5 25.3 56398
Wood 1.2 0.5 0.2 0.0 8.7 1.7 0.6 0.2 1845
Other 9.9 3.9 1.9 0.4 13.4 6.4 3.6 0.3 2232
Not specified 54.3 29.7 16.3 0.1 54.3 37.0 25.2 0.0 92
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 69 -
Table F-11: Incidence, depth and severity of poverty by source of drinking water (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Piped in dwelling 1.6 0.4 0.1 3.3 4.0 1.1 0.4 4.2 106214
Piped on site 9.9 3.0 1.2 10.5 21.6 6.8 3.1 11.5 54324
Neighbor's tap 19.2 5.7 2.5 7.5 38.2 12.5 5.9 7.5 20156
Public tap 19.0 5.4 2.3 35.5 39.0 12.4 5.6 36.4 95600
Water carrier/tanker 16.6 5.1 2.2 0.8 26.0 10.0 4.9 0.6 2358
Private bore hole 23.2 5.3 1.9 4.1 44.9 13.7 5.8 3.9 8958
Communal bore hole 22.8 6.8 3.0 11.4 43.4 14.8 7.0 10.8 25536
Protected well 20.3 5.0 1.9 4.3 43.5 13.3 5.6 4.6 10723
Spring 10.9 0.7 0.0 0.1 20.2 5.9 1.8 0.1 496
Flowing water 34.3 11.0 5.1 11.7 51.9 20.6 10.7 8.9 17514
Rain water tank 35.3 4.1 1.0 0.5 62.1 17.0 6.0 0.5 762
Unprotected well 19.1 5.6 2.5 7.5 39.7 12.4 5.8 7.8 20234
Dam/pool/stagnant
water
17.5 3.7 1.1 2.4 40.9 11.6 4.4 2.8 7077
Other 17.3 5.5 2.2 0.5 30.9 10.5 5.1 0.4 1340
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-12: Incidence, depth and severity of poverty by toilet facilities (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Flush/sewer 2.3 0.7 0.3 5.7 5.9 1.7 0.7 7.3 127114
Flush/septic tank 3.8 0.4 0.1 0.7 10.3 2.3 0.7 0.9 9276
Pit latrine/VIP 8.1 1.8 0.6 2.2 16.8 5.1 2.0 2.3 14091
Pit latrine/ no
ventilation
16.6 4.7 2.0 5.6 33.6 10.5 4.8 5.6 17205
Bucket 24.8 7.6 3.2 2.3 40.5 15.5 7.6 1.9 4702
Other 10.2 2.0 0.5 0.2 22.1 7.0 2.6 0.2 859
Bush 21.6 6.2 2.7 83.4 42.3 13.9 6.4 81.8 197802
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 70 -
Table F-13: Incidence, depth and severity of poverty by ownership/access to cattle (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 12.2 3.0 1.2 29.7 26.5 7.9 3.3 32.4 125325
Does not own,
but has access
19.5 5.9 2.7 10.0 39.1 13.1 6.2 10.0 26259
Neither owns nor
has access
14.0 4.2 1.8 60.1 26.8 9.0 4.3 57.5 219831
Not stated 26.2 11.4 5.9 0.1 26.2 15.8 10.0 0.1 263
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-14: Incidence, depth and severity of poverty by ownership/access to goats (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 13.4 3.1 1.2 37.9 30.2 8.7 3.5 42.7 145027
Does not own,
but has access
9.9 2.6 0.9 2.6 22.1 6.6 2.8 2.9 13305
Neither owns nor
has access
14.3 4.5 2.1 59.2 26.1 9.2 4.5 54.1 212703
Not stated 28.0 13.0 6.8 0.4 40.7 19.9 11.7 0.3 643
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-15: Incidence, depth and severity of poverty by ownership/access to field for crops (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 19.5 5.6 2.5 35.4 38.1 12.6 5.9 34.7 93332
Does not own,
but has access
13.7 3.3 1.2 28.9 32.9 9.1 3.7 34.7 108232
Neither owns nor
has access
10.7 3.4 1.5 35.2 18.3 6.7 3.4 30.2 169077
Not stated 23.8 5.5 2.1 0.5 35.5 13.1 5.9 0.4 1037
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 71 -
Table F-16: Incidence, depth and severity of poverty by ownership/access to radio (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 11.4 3.1 1.3 59.0 23.6 7.4 3.3 61.2 265491
Does not own,
but has access
20.6 6.1 2.8 19.6 39.9 13.2 6.3 19.0 48863
Neither owns nor
has access
19.3 6.0 2.7 21.3 35.5 12.4 6.0 19.6 56819
Not stated 13.7 6.5 3.3 0.1 30.9 11.1 6.0 0.2 505
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-17: Incidence, depth and severity of poverty by ownership/access to plough (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 15.4 4.0 1.7 25.3 36.6 10.5 4.4 30.0 84033
Does not own,
but has access
19.7 5.4 2.3 18.5 40.1 12.8 5.8 18.8 48226
Neither owns nor
has access
12.0 3.6 1.6 55.8 21.9 7.6 3.6 50.9 238230
Not stated 17.4 5.8 2.2 0.4 27.0 10.7 5.3 0.3 1189
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-18: Incidence, depth and severity of poverty by ownership/access to telephone (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Owns 1.9 0.4 0.1 4.6 4.9 1.2 0.4 5.9 124528
Does not own,
but has access
15.5 4.1 1.6 37.4 33.5 10.1 4.4 40.4 123603
Neither owns nor
has access
24.1 7.4 3.3 57.7 44.6 15.5 7.5 53.4 122603
Not stated 20.0 6.8 2.6 0.4 41.3 13.9 6.5 0.4 944
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 72 -
Table F-19: Incidence, depth and severity of poverty by energy source for cooking (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Electricity
from mains
1.3 0.4 0.2 2.6 3.5 0.9 0.4 3.6 106048
Electricity from
generator
.. 0.0 0.0 .. 9.0 1.6 0.3 0.1 1097
Solar energy .. 0.0 0.0 .. 17.1 4.7 1.3 0.0 69
Gas 4.1 1.0 0.3 1.7 10.9 2.8 1.1 2.3 21691
Paraffin 6.6 1.3 0.5 2.1 16.2 4.1 1.6 2.6 16430
Wood 21.2 6.1 2.6 91.6 41.6 13.7 6.3 89.7 221380
Coal 19.4 9.7 6.1 0.2 37.7 15.3 9.4 0.2 640
Animal dung 22.4 4.4 1.1 1.7 34.5 11.4 4.6 1.3 3817
Other .. 0.0 0.0 .. 39.1 8.3 1.8 0.1 138
None .. 0.0 0.0 .. 22.3 4.4 0.9 0.1 369
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
Table F-20: Incidence, depth and severity of poverty by energy source for lighting (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
Electricity
from mains
2.8 0.8 0.3 7.3 6.5 1.9 0.9 8.5 132916
Electricity
from generator
3.3 0.5 0.1 0.2 6.2 1.9 0.7 0.2 2537
Solar energy 3.0 0.6 0.3 0.1 6.3 1.5 0.7 0.1 1623
Gas 7.8 2.2 0.9 0.1 18.3 6.4 2.6 0.1 677
Paraffin 12.5 3.0 1.2 13.7 30.5 8.4 3.4 16.8 56269
Wood 34.3 11.4 5.4 15.9 61.1 22.9 11.4 14.2 23775
Candles 20.1 5.5 2.3 56.0 39.2 12.6 5.7 54.6 142735
Other 30.8 10.6 5.5 5.6 50.9 19.3 10.3 4.6 9319
None 33.4 12.3 6.1 1.1 60.5 22.9 11.9 1.0 1745
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 73 -
Table F-21: Urbanisation, and incidence of poverty by region and urban/rural (%)
Urbanisation
levels
Severely poor: Lower
bound poverty line
(N$184.56 per adult
equivalent)
Poor: Upper bound
poverty line
(N$262.45 per adult
equivalent)
Urban Rural Urban
Poor
Rural Poor Urban
Poor
Rural
Poor
Number of
households
Caprivi 27.7 72.3 3.1 16.1 12.0 34.9 18607
Erongo 83.8 16.2 2.8 15.3 6.9 27.9 27713
Hardap 39.5 60.5 17.2 24.9 24.3 37.1 16365
Karas 53.8 46.2 8.5 17.3 17.7 26.9 15570
Kavango 20.1 79.9 19.7 41.0 32.8 62.4 32354
Khomas 92.5 7.5 1.7 10.6 5.3 18.3 64918
Kunene 32.6 67.4 16.0 11.7 28.0 20.5 13365
Ohangwena 2.0 98.0 3.7 19.6 7.9 45.5 37854
Omaheke 24.3 75.7 16.7 17.7 24.4 31.9 13347
Omusati 1.8 98.2 7.2 12.9 10.4 31.4 39248
Oshana 41.2 58.8 4.4 10.1 9.9 26.5 31759
Oshikoto 13.0 87.0 10.6 17.5 19.9 43.9 31871
Otjozond-
jupa
50.7 49.3 12.7 18.9 22.7 33.0 28707
Namibia 40.5 59.5 6.0 19.1 12.0 38.2 371678
Table F-22: Urbanisation, and poverty shares by region and urban/rural (%)
Distribution of
rural/urban
households
Severely poor: Lower
bound poverty line
(N$184.56 per adult
equivalent)
Poor: Upper
bound poverty line
(N$262.45 per
adult equivalent)
Urban Rural Urban
Poor
Rural
Poor
Urban
Poor
Rural
Poor
Caprivi 3.4 6.1 1.8 5.1 3.4 5.6
Erongo 15.4 2.0 7.2 1.6 8.9 1.5
Hardap 4.3 4.5 12.2 5.9 8.7 4.4
Karas 5.6 3.3 7.8 2.9 8.2 2.3
Kavango 4.3 11.7 14.1 25.1 11.7 19.1
Khomas 39.9 2.2 11.1 1.2 17.4 1.1
Kunene 2.9 4.1 7.6 2.5 6.7 2.2
Ohangwena 0.5 16.8 0.3 17.2 0.3 20.0
Omaheke 2.2 4.6 6.0 4.2 4.4 3.8
Omusati 0.5 17.4 0.5 11.8 0.4 14.3
Oshana 8.7 8.5 6.3 4.5 7.1 5.9
Oshikoto 2.8 12.5 4.8 11.5 4.6 14.4
Otjozond-
jupa
9.7 6.4 20.3 6.3 18.2 5.5
Namibia 100.0 100.0 100.0 100.0 100.0 100.0
- 74 -
Table F-23: Poverty incidence and shares by region (rank)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
Urban
Poor
Urban
Poverty
share
Rural
Poor
Rural
Poverty
share
Urban
Poor
Urban
Poverty
share
Rural
Poor
Rural
Poverty
share
Caprivi 11 11 8 7 8 11 5 6
Erongo 12 7 9 12 12 4 9 12
Hardap 2 3 2 6 4 5 4 8
Karas 7 5 7 10 7 6 10 10
Kavango 1 2 1 1 1 3 1 2
Khomas 13 4 12 13 13 2 13 13
Kunene 4 6 11 11 2 8 12 11
Ohangwena 10 13 3 2 11 13 2 1
Omaheke 3 9 5 9 3 10 7 9
Omusati 8 12 10 3 9 12 8 4
Oshana 9 8 13 8 10 7 11 5
Oshikoto 6 10 6 4 6 9 3 3
Otjozondjupa 5 1 4 5 5 1 6 7
Table F-24: Incidence, depth and severity of poverty by highest level of educational
attainment of the household head (%)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
P(0)
incidence
(P1)
depth
(P2)
severity
Poverty
share
Number of
households
No formal
education
26.7 8.1 3.6 46.1 50.0 17.2 8.2 43.1 88375
Primary
education
17.7 5.0 2.1 40.3 35.5 11.4 5.2 40.4 116545
Secondary
education
5.1 1.2 0.5 12.6 12.6 3.5 1.4 15.6 126932
Tertiary
education
0.1 0.0 0.0 0.0 0.4 0.1 0.0 0.2 36980
Don't know 24.4 5.3 2.2 0.6 42.4 13.3 5.8 0.5 1327
Missing 12.4 2.1 0.5 0.4 15.0 5.6 2.3 0.2 1518
Namibia 13.8 3.9 1.7 100.0 27.6 8.9 4.1 100.0 371678
- 75 -
Table F-25: Mean distances to facilities and services by poverty incidence (minutes and kilometers)
Severely poor: Lower bound poverty line
(N$184.56 per adult equivalent)
Poor: Upper bound poverty line
(N$262.45 per adult equivalent)
Not Poor Poor Total Not Poor Poor Total
Distance in minutes
Drinking water 4.60 6.63 4.88 4.05 7.04 4.88
Hospital or clinic 13.51 12.57 13.38 12.57 15.49 13.38
Public transport 7.81 7.37 7.75 6.97 9.78 7.75
Local shop/market 8.64 7.99 8.55 8.40 8.93 8.55
Primary school 10.48 8.87 10.26 10.35 10.04 10.26
High school 14.85 13.48 14.66 14.23 15.80 14.66
Combined school 15.83 8.84 14.86 16.32 11.03 14.86
Police station 14.48 12.61 14.23 13.80 15.34 14.23
Post office 13.92 11.56 13.59 13.45 13.96 13.59
Magistrate court 13.69 8.50 12.98 14.01 10.28 12.98
Traditional court 7.27 6.38 7.15 6.99 7.58 7.15
Mobile clinic 1.06 1.39 1.11 0.88 1.70 1.11
Distance in kilometers
Drinking water 0.62 1.09 0.68 0.53 1.08 0.68
Hospital or clinic 10.89 13.52 11.25 10.77 12.52 11.25
Public transport 6.31 9.11 6.69 6.19 8.02 6.69
Local shop/market 6.89 8.86 7.16 6.86 7.96 7.16
Primary school 6.89 8.07 7.05 6.91 7.44 7.05
High school 24.49 34.53 25.87 23.87 31.14 25.87
Combined school 30.51 40.29 31.86 31.09 33.89 31.86
Police station 15.10 27.99 16.88 14.13 24.11 16.88
Post office 19.81 35.41 21.96 18.73 30.45 21.96
Magistrate court 30.88 48.59 33.33 29.04 44.60 33.33
Traditional court 14.78 4.97 13.42 16.25 6.02 13.42
Mobile clinic 0.30 0.23 0.29 0.31 0.25 0.29
- 76 -
Table F-26: Table Poverty share by ownership/access to household assets (%)
Severely poor Poor Non-poor All Namibia
Owns radio 59.0 61.2 75.3 71.4
Does not own but access to radio 19.6 19.0 10.9 13.1
Owns a telephone/cellphone 4.6 5.9 44.0 33.5
Does not own but access to telephone/cellphone 37.4 40.4 30.5 33.3
Owns a motor vehicle 1.1 1.6 24.9 18.5
Does not own but access to motor vehicle 24.4 26.4 28.8 28.1
Owns a television 3.8 5.2 38.3 29.1
Does not own but access to television 7.6 8.1 11.1 10.3
Owns a refrigerator 2.7 4.6 40.1 30.3
Does not own but access to refrigerator 4.5 5.4 6.7 6.3
Owns a tape recorder 8.3 10.5 34.6 27.9
Does not own but access to tape recorder 6.4 5.9 6.6 6.4
Owns a HiFi 2.9 4.2 32.1 24.4
Does not own but access to HiFi 6.5 6.0 6.8 6.6
Owns a freezer 1.1 1.8 25.9 19.3
Does not own but access to freezer 2.8 3.8 5.8 5.2
Owns a camera 1.5 1.8 18.4 13.8
Does not own but access to camera 8.0 9.3 9.9 9.7
Owns a bicycle 8.1 9.9 17.8 15.6
Does not own but access to bicycle 4.0 5.7 7.3 6.9
Owns a sewing/knitting machine 8.5 9.7 18.3 15.9
Does not own but access to sewing/knitting machine 4.5 6.3 6.2 6.2
Owns a VCR/DVD 0.4 0.7 17.2 12.6
Does not own but access to VCR/DVD 2.1 2.3 4.9 4.2
Owns a washing machine 0.8 1.1 18.7 13.9
Does not own but access to washing machine 0.9 1.2 2.7 2.3
Does not own but access to microwave 0.6 0.6 2.2 1.8
Owns a satellite dish 0.1 0.2 11.3 8.3
Does not own but access to satellite dish 1.0 1.2 4.7 3.8
Owns a computer .. 0.1 7.4 5.4
Does not own but access to computer 0.5 0.6 8.7 6.5
Owns a Internet service .. 0.1 3.9 2.8
Does not own but access to Internet service 0.2 0.3 7.7 5.6
Owns a canoe/boat 1.4 2.1 1.3 1.5
Does not own but access to canoe/boat 2.2 2.4 2.1 2.1
Owns a motor cycle/scooter 0.2 0.2 1.4 1.1
Does not own but access to motor cycle/scooter 0.1 0.3 1.1 0.9
Owns a motorboat .. .. 0.3 0.3
Does not own but access to motorboat 0.1 0.2 0.6 0.5
Does not own but access to microwave 0.6 0.6 2.2 1.8
Owns a satellite dish 0.1 0.2 11.3 8.3
- 77 -
Table F-27: Poverty share by ownership/access to agricultural assets (%)
Severely poor Poor Non-poor All Namibia
Owns grazing land 2.0 2.3 5.5 4.7
Does not own, but access to grazing land 58.7 62.9 47.4 51.7
Owns field for crops 35.4 34.7 21.5 25.1
Does not own, but access to field for crops 28.9 34.7 27.0 29.1
Owns poultry 58.5 63.5 43.0 48.6
Does not own, but access to poultry 1.4 1.4 2.7 2.3
Owns goats 37.9 42.7 37.6 39.0
Does not own, but access to goats 2.6 2.9 3.9 3.6
Owns cattle 29.7 32.4 34.2 33.7
Does not own, but access to cattle 10.0 10.0 5.9 7.1
Owns plough 25.3 30.0 19.8 22.6
Does not own, but access to plough 18.5 18.8 10.7 13.0
Owns wheelbarrow 9.2 10.8 22.2 19.1
Does not own, but access to wheelbarrow 10.9 12.7 12.4 12.5
Owns donkey/mule 19.4 20.6 16.1 17.3
Does not own, but access to donkey/mule 4.3 4.9 3.5 3.9
Owns donkey cart/ox cart 9.8 10.1 7.6 8.3
Does not own, but access to donkey cart/ox cart 8.4 9.7 6.8 7.6
Owns pig 14.8 18.5 12.7 14.3
Does not own, but access to pig 0.3 0.7 1.2 1.1
Owns tractor 0.5 0.3 1.7 1.3
Does not own, but access to tractor 10.7 12.6 11.8 12.0
Owns grinding mill 0.2 0.2 2.5 1.9
Does not own, but access to grinding mill 5.4 8.0 10.4 9.8
Owns sheep 3.9 3.7 7.5 6.4
Does not own, but access to sheep 1.2 1.1 1.4 1.3
Owns horse 4.6 3.9 6.0 5.4
Does not own, but access to horse 1.1 1.6 1.2 1.3
Owns ostrich 0.1 0.1 0.5 0.4
Does not own, but access to ostrich 0.3 0.2 0.2 0.2
- 78 -
Table F-28: Poverty share by lack of ownership/access to household assets (%)
Severely poor Poor Non-poor All Namibia
Radio 21.3 19.6 13.6 15.3
Telephone/Cellphone 57.7 53.4 25.2 33.0
Motor vehicle 74.0 71.6 46.0 53.1
Stove, gas or electric 88.9 86.6 42.8 54.9
Television 88.1 86.3 50.3 60.3
Refrigerator 92.5 89.8 53.0 63.1
Tape Recorder 84.8 83.3 58.4 65.3
HiFi 90.0 89.4 60.7 68.7
Freezer 95.9 94.1 68.1 75.3
Camera 90.0 88.5 71.3 76.1
Bicycle 87.5 84.1 74.2 76.9
Sewing/knitting machine 86.5 83.7 75.1 77.5
VCR/DVD 97.1 96.6 77.4 82.7
Washing machine 97.8 97.4 78.4 83.6
Microwave 98.9 98.8 81.4 86.2
Satellite dish 98.5 98.2 83.5 87.6
Computer 99.2 99.0 83.3 87.7
Internet service 99.4 99.3 88.1 91.2
Canoe/boat 96.1 95.2 96.2 95.9
Motor cycle/scooter 99.2 99.1 97.0 97.6
Motorboat 99.5 99.5 98.5 98.8
Table F-29: Poverty share by lack of ownership/access to agricultural assets (%)
Severely poor Poor Non-poor All Namibia
Grazing land 39.1 34.5 46.9 43.5
Field for crops 35.2 30.2 51.3 45.5
Poultry 39.8 35.1 54.2 48.9
Goat 59.2 54.1 58.4 57.2
Cattle 60.1 57.5 59.8 59.1
Plough 55.8 50.9 69.1 64.1
Wheelbarrow 79.5 76.1 64.9 68.0
Donkey/mule 75.9 74.3 80.2 78.6
Donkey cart/ox cart 81.3 79.8 85.2 83.7
Pig 84.6 80.6 85.8 84.4
Tractor 88.4 86.8 86.1 86.3
Grinding mill 93.7 91.3 86.7 88.0
Sheep 94.7 95.0 91.0 92.1
Horse 93.7 94.2 92.6 93.0
Ostrich 99.1 99.3 98.9 99.0
- 79 -
ANNEX G: Multivariate analysis
This Annex presents the methodology and more detailed results from the multivariate
analysis introduced in Section 5. Two types of regression models are presented. First an
Ordinary Least Squares model is estimated using monthly household expenditure as the
dependent variable (i.e. the variable to be explained) and a series of socio-economic char-
acteristics as independent variables (i.e. variables that explain variation in the dependent).
This model is specified as:
ln y
j
= ²x
j
+ µ
j
Where yj is total monthly adult equivalent expenditure of household j in Namibian $; xj is a
set of exogenous household characteristics or other determinants, and µj is a random error
term. The dependent variable is logarithmically transformed which means that the coeffi-
cients of the independents can be interpreted as partial effects in percentage terms.
The second model is a binary logistic model where the dependent is the categorical variable
of poverty status which takes a value of 1 if the household is classified as poor and a value
of 0 if the household is classified as non-poor. This model takes the form:
Prob(poor=1) = (exb)/(1+ exb)
Where x² = ±+²
1
x
1
+²
2
x
2
+&+²
k
x
k
The selection of explanatory variables included in the models were based on the informa-
tion from the poverty profile and other areas of interest to policymakers. The variables
should ideally be exogenous to the level of welfare. In fact, most variables at some point in
time end up being determined to some extent by the welfare of the households (except a
few such as age and gender) so the definition of exogenous refers to the short term. For in-
stance variables such as housing standards are not included since they are likely to be a di-
rect function of current levels of welfare.
When using categorical variables (e.g. education which is expressed as primary, secon-
dary, tertiary and no formal education), a reference category is selected as a default
and omitted from the regression. The resulting parameter estimates should be interpreted in
relation to the default category (in the case of education the default category is secondary
education). In principle, any category could be used as a default but those selected in this
analysis were chosen to meaningfully represent the variable, i.e. they include a large num-
ber of observations and are easy to interpret. It should be emphasized that the models are
only able to provide correlations from which no inference of causation can be made. In-
stead the model can assist in testing the strength of relationships in the NHIES data that
have been to shown to be causal elsewhere, e.g. between the level of education and welfare
or between fertility and poverty status.
- 80 -
The regressions were all run using SPSS software and the sampling weights were rebased
to N=9801 in order to get meaningful significance levels.
Determinants of household consumption
When using categorical variables, a reference category is selected as a default and omitted
from the regression. The results from the regression using the national data are reported in
Table G-1 and for each of the 13 administrative regions in the country in Table G-2. Levels
of significance at levels greater than 1, 5 and 10 percent are indicated.
As would be expected given the results of the poverty profile, there is a strongly inverse
relationship between adult equivalent adjusted household expenditure and the size of the
household. Increasing the size of the household reduces adult equivalent household expen-
diture by 23.9 percent when all other factors are controlled for. This result also holds for all
the regions individually.
The analysis confirms the gender dimensions of household levels of welfare. Female-
headed households have total consumption expenditures that are lower by 4.9 percent com-
pared to the default category of male-headed households. In other words, when comparing
a household headed by a female and one headed by a male, the former will have consump-
tion levels that are around 5 percent lower even when controlling for differences in the
level of education, number of people in the household, location and so on. This is evidence
that there is a gender aspect to poverty in Namibia as is often found in other developing
countries. There are some sizeable region-specific differences when it comes to gender-
inequality. The biggest difference between male- and female-headed households is in the
regions of Omaheke, Oshikoto, Khomas and Oshana, where household expenditures among
female-headed households are lower by more than 9 percent compared to male-headed
households and controlling for all other factors. As would be expected from the analysis of
the poverty profile, household consumption expenditures also increase with the age of the
head of household but at a decreasing pace indicated by the negative coefficient of the
squared age variable. It should be noted that just like under the poverty profile, the results
here refer to the sex and age of the head of household and does not take into account issues
related to intra-household inequality.
Having one or more children in the household reduces adult equivalent consumption by 12
percent compared to households without any children and holding other factors, including
household size, constant. This relationship is statistically significant in all but two regions
and strongest in Ohangwena, where having one or more children in the household lowers
the adult equivalent consumption expenditure by 24.5 percent compared to households
without any children. Under a different specification of the model, a dummy variable was
included to test for the relationship between the presence of an orphan in the household and
consumption levels. This relationship proved insignificant once other factors, including the
presence of children in the household, are controlled for.
The analysis confirms the great regional differences in levels of consumption expenditure
among households. Rural households also have lower levels of consumption expenditure
compared to the urban default controlling for all other factors and this result holds for all
- 81 -
the 11 regions where the relationship is significant. In rural Oshikoto and rural Hardap, ex-
penditure levels are lower by more than 20 percent compared to the urban areas of those
regions and controlling for other factors related to the household. The strongest impacts on
consumption expenditure of households come from the education variables. In households
where the head has primary education as the highest level of education or has no formal
education at all, the monthly consumption levels are lower by 19.8 and 24.4 percent, re-
spectively compared to households where the head has attained a secondary level of educa-
tion. Conversely, in households where the head has attained a tertiary education, the con-
sumption levels are higher by 26.6 percent compared to household heads with a secondary
education. The correlation between education and expenditure levels is strongly significant
in all regions of the country.
Having a pension as the main source of income reduces consumption expenditure by 4.6
percent compared to all other sources of income including wages, income from subsistence
farming and non-farming business activities. In Karas and Kunene, having a pension as the
main source of income is associated with lower consumption expenditure of 11 and 19 per-
cent, respectively compared to households with other sources of income. On the other
hand, owning or having access to field for grazing increases household consumption by 8.5
percent. The variables reflecting distances to public services and facilities are somewhat
ambiguous. Expenditure levels increase with distance to hospital/clinic and shop/market
but decrease with distance to police station. The latter effect could be a sign that better-off
neighbourhoods are better policed compared to less well-off areas or, if one assumes that
more crimes take place further away from police presence, then the effect could be inter-
preted as the negative impact of crime on household incomes and consumption. However,
the model is not able to determine such a causal links. Moreover, in some regions the corre-
lation is negative e.g. in Omusati and Kavango, while in yet other regions e.g. Omaheke
and Erong the relationship is positive (i.e. the farther away from a police station the house-
hold is located, the higher the expenditure levels).
Consumption expenditures are lowest in Caprivi and Kavango compared to Ohangwena,
which is the default category for the regions, when controlling for the effects of other vari-
ables. The regions of Karas, Hardap and Oshikoto also have lower levels of consumption
expenditure. On the other hand, Khomas, Omusati and Oshana have higher levels of con-
sumption expenditure. This may seem to contradict the results from the poverty profile,
which showed that Ohangwena ranked second highest in terms of both levels of poverty
and poverty share. The reason for the change in ranking is that the multivariate analysis
controls for other factors that determine poverty status and shows the strength of the effects
that are attributable to the region per se.
This way, the results show that when holding constant all other characteristics that are
thought to influence income and consumption levels, for example education levels, age,
number of children in the household and so on, a household in Caprivi is likely to be poorer
than a household living in any other region of the country. Likewise, a household in
Khomas is more likely to have a higher level of income or consumption than in any other
region. The regression analysis also included the language variables as explanatory, and the
coefficients are significant for all categories expect for households where the main lan-
- 82 -
guage is Setswana. Again, some interesting shifts occur compared to the poverty profile. In
households where Afrikaans is the main language, total consumption is higher by 19.8 per-
cent compared to the default category, which is Oshiwambo, and households where Ger-
man and English consumption is higher by 11.3 and 10.5 percent respectively. On the other
hand, households where the main languages spoken are Khoisan, Rukavango and espe-
cially Nama/Damara, total consumption levels are lower (than the default category), again
holding constant all other factors that are included in the model.
- 83 -
Table G-1: Results of OLS regression
Dependent Variable: Ln(adult equivalent monthly expenditure)
N=9801
Adj R Squared: 0.629
Standardized ² -
coefficients
T
(Constant) 114.934
Household size -.239*** -29.193
Age of head of household .219*** 6.287
Age of head of household (squared) -.182*** -5.264
Female (=1; Male =0) -.049*** -7.043
Child younger than 16 (=1; no child =0) -.120*** -15.126
Widow/widower (=1; other marital status =0) .018** 2.435
Rural (=1; Urban=0) -.116*** -11.644
Distance to hospital/clinic (km) .047*** 4.892
Distance to shop/market (km) .028*** 3.344
Distance to police station (km) -.042*** -4.772
Owns or has access to field for crops (=1; Neither
owns nor has access=0)
-.002 -.220
Owns or has access to field for grazing (=1; Neither
owns nor has access=0)
.085*** 9.996
Pension (=1; Other source of income=0) -.046*** -6.470
Education dummies (default: secondary education)
Primary education -.198*** -26.109
Tertiary education .266*** 37.368
No formal education -.244*** -29.393
Regional dummies (default: Ohangwena)
Caprivi -.054*** -3.537
Erongo .022** 2.308
Hardap -.041*** -4.672
Karas -.034*** -4.128
Kavango -.044*** -2.956
Khomas .085*** 6.948
Kunene .012 1.488
Omaheke .003 .386
Omusati .017** 2.047
Oshana .053*** 6.344
Oshikoto -.025*** -3.052
Otjozondjupa -.012 -1.204
Language dummies (default: Oshiwambo)
Khoisan -.022*** -3.330
Caprivian .038** 2.573
Otjiherero .046*** 5.570
Rukavango -.023* -1.649
Nama/damara -.036*** -4.267
Setswana .009 1.358
Afrikaans .198*** 23.643
German .113*** 17.168
English .105*** 15.412
Other .037*** 5.763
* = p < 0.1; ** = p < 0.05 *** = p < 0.01.
- 84 -
Table G-2: Results of OLS regression by region
Dependent Variable: Ln(adult equivalent monthly expenditure)
N= 731 545 640 645 714 1160 478 725 495 964 1005 998 701
Adj R Squared: 0.457 0.503 0.493 0.499 0.535 0.556 0.515 0.506 0.486 0.395 0.478 0.484 0.514
Caprivi Erongo Hardap Karas Kavango Khomas Kunene Ohangwena Omaheke Omusati Oshana Oshikoto Otjzondjupa
Standardised ²-coefficients
Household size -0.312*** -0.220*** -0.284*** -0.329*** -0.349*** -0.272*** -0.471*** -0.281*** -0.269*** -0.243*** -0.237*** -0.209*** -0.302***
Age of household head 0.148 0.448*** 0.175 0.482** 0.255* 0.302*** 0.182 -0.033 0.705*** 0.083 0.081 0.185 0.465***
Age of household head (sq) -0.028 -0.348** -0.008 -0.409** -0.232* -0.124 0.020 -0.067 -0.513** -0.050 -0.037 -0.204 -0.415***
Female 0.040 -0.076*** 0.005 -0.052 -0.044 -0.093*** -0.020 -0.058* -0.094** -0.044 -0.091*** -0.093*** -0.086***
Child in household -0.178*** -0.156*** -0.136*** -0.055 -0.073*** -0.041* -0.073 -0.245*** -0.166*** -0.188*** -0.086*** -0.188*** -0.160***
Rural -0.173*** -0.170*** -0.207*** -0.106** -0.115*** -0.131*** 0.028 -0.063*** -0.187*** 0.022 -0.226*** -0.196*** -0.098***
Widow 0.015 -0.031 0.042 0.005 0.044 0.061*** -0.016 0.005 0.008 0.020 0.007 -0.019 0.012
Hospital/clinic -0.020 0.025 -0.045 0.081 -0.054* 0.126** 0.072 -0.039 -0.029 -0.085** -0.018 -0.046 0.141***
Shop/market 0.036 -0.070 0.179** 0.037 0.085** 0.001 0.019 0.043 0.146*** 0.037 0.008 0.046 -0.027
Police station -0.039 0.149** 0.070 -0.105 -0.180*** -0.053 0.021 0.026 0.170*** -0.221*** 0.011 -0.063 -0.039
Grazing 0.115*** -0.087** 0.120*** 0.141*** 0.210*** 0.010 0.115*** 0.050* 0.133*** 0.003 0.133*** 0.073** 0.097**
Crops 0.007 0.143*** -0.012 -0.035 -0.066** -0.074*** -0.034 -0.046* 0.055 -0.114*** -0.024 0.039 0.041
Primary education -0.228*** -0.260*** -0.325*** -0.231*** -0.278*** -0.234*** -0.249*** -0.202*** -0.308*** -0.178*** -0.232*** -0.266*** -0.299***
Tertiary education 0.289*** 0.404*** 0.247*** 0.341*** 0.297*** 0.437*** 0.178*** 0.281*** 0.197*** 0.208*** 0.304*** 0.295*** 0.193***
No formal -0.271*** -0.213*** -0.420*** -0.280*** -0.330*** -0.249*** -0.368*** -0.269*** -0.389*** -0.243*** -0.269*** -0.329*** -0.448***
Pension -0.062 0.036*** -0.113** -0.109** -0.042 -0.055*** -0.193*** -0.020 -0.076* -0.068*** 0.018 -0.018 0.002
* = p < 0.1; ** = p < 0.05 *** = p < 0.01.
- 85 -
Determinants of household poverty status
The second type of multivariate analysis conducted on the data makes use of the new pov-
erty line definition. Monthly consumption is replaced as the dependent variable with a vari-
able of two categories representing the poverty status of households; 1=poor and 0=non-
poor using the poverty line defined for this report. The independent variables are the same
as under the previous model. The two methods are quite similar but the properties of the
coefficients differ. The regression coefficients of the logistic model are converted to odds
ratios, which each signify the probability of the household with that characteristic being
poor when controlling for all other factors. Results are reported in Table G-3.
The highest odds ratio is for no formal education of the head of household. These house-
holds have an odds ratio of 4.2. In other words, households where the head has no formal
education are more than four times as likely to be classified as poor compared to house-
holds where the head has a secondary education and controlling for all other factors.
Households where primary education is the highest level of education attained by head of
household are also more likely to be poor. The analysis further shows that households in
rural areas have an odds ratio of 1.97, which means that they are 97 percent more (almost
twice as) likely to be poor compared to urban households and holding all other factors con-
stant.
Additional factors contribute to the probability of household poverty. Having a child
younger than 16 in the household make it 1.77 times (or 77 percent) more likely to be poor
compared to households without any children. Households where pension is the main
source of income are 1.74 times more likely to be poor than households that rely on other
main sources of income. Female-headed households are 1.18 times as likely to be poor
compared to male-headed households. Several regional variables, Caprivi, Kavango and
Oshikoto, also have odds ratios higher than one, which indicates that households residing
in these regions are more likely to be poor, compared to households residing in Ohangwena
(the default category) and holding all other variables constant.
Conversely, several factors have odds ratios below 1, which means that the probabilities
shift towards the household being less likely than the default category to be classified as
poor. The most important of these factors is tertiary education. An odds ratio of 0.019 im-
plies that if the household head has a tertiary education, it is 50 times less likely to be poor
compared to a household where the head has a secondary education. Moreover, households
residing in the regions of Erongo, Kunene, Oshana and Khomas are half as likely to be
poor compared to those in Ohangwena when all other factors are controlled for.
- 86 -
Table G-3: Results of Binary Logistic regression for Namibia
Dependent variable: Poverty status (poor=1, non-poor=0)
N=9801
Pseudo R Squared: 0.427
Hosmer Lemeshow goodness of fit: Ç2=10.338; p=.242
² Exp (²)
Household size 0.214*** 1.239
Age of head of household -0.015* 0.985
Age of head of household (squared) 0.000* 1.000
Female (=1; Male =0) 0.172*** 1.188
Child younger than 16 (=1; no child =0) 0.571*** 1.769
Widow/widower (=1; other marital status =0) -0.189** 0.828
Rural (=1; Urban=0) 0.678*** 1.969
Distance to hospital/clinic (km) -0.007*** 0.993
Distance to shop/market (km) -0.003 0.997
Distance to police station (km) 0.006*** 1.006
Owns or has access to field for grazing (=1; Neither owns
nor has access=0) -0.549*** 0.578
Owns or has access to field for crops (=1; Neither owns nor
has access=0) -0.055 0.947
Pension (=1; Other source of income=0) 0.556*** 1.744
Education dummies (default: secondary education)
Primary education 1.028*** 2.796
Tertiary education -3.944*** 0.019
No formal education 1.436*** 4.204
Regional dummies (default: Ohangwena)
Caprivi 0.907** 2.476
Erongo -0.561*** 0.571
Hardap 0.208 1.231
Karas 0.061 1.063
Kavango 0.545** 1.725
Khomas -0.871*** 0.419
Kunene -0.592*** 0.553
Omaheke 0.097 1.102
Omusati -0.453*** 0.636
Oshana -0.783*** 0.457
Oshikoto 0.286** 1.331
Otjozondjupa 0.067 1.069
Language dummies(default: Oshiwambo)
Khoisan 0.478** 1.614
Caprivian -0.964** 0.382
Otjiherero -0.488*** 0.614
Rukavango 0.120 1.128
Nama/damara 0.451*** 1.570
Setswana -1.102** 0.332
Afrikaans -0.818*** 0.441
German -17.156 0.000
English -1.452 0.234
Other 0.117 1.125
Constant -2.988*** 0.050
* = p < 0.1; ** = p < 0.05 *** = p < 0.01.
- 87 -
ANNEX H: Measures of inequality and polarisation
This Annex supplements Section 6 of the main report by presenting a deeper and more
formal analysis of inequality and polarisation in Namibia. The presentation is based on Du-
clos and Araar (2006) and the same notations for the mathematical expressions are used.
The three main inequality indices included are those of the S-Gini class, the Generalised
Entropy class and the Atkinson. The two measures of polarisation are the Wolfson index
and the Duclos, Esteban and Ray (DER) index.9
First, the Lorenz curve, which plots cumulative share of expenditure against the cumulative
share of households ranked by expenditure can be defined as:
+=+
+
=
p
0
1
0
p
0
dq)q(Q
µ
1
dq)q(Q
dq)q(Q
)p(L
The numerator sums the consumption expenditure of the bottom percentile, p. The denomi-
nator sums the consumption expenditure of all households. Under perfect equality expendi-
ture shares and the share of households are the same L(p) = p and so aggregating the dis-
tance p L(p) over the entire expenditure distribution yields the most common measure of
inequality used in Namibia:
The Gini index of inequality = +1
0
dp))p(Lp(2
This measure applies equal weights in the aggregation of p L(p). However, it is possible
to define percentile dependent weights º(p) and apply these to the measured distances in
order to reflect that society is concerned more with inequality among the poorest. Typically
such weights are defined as:
2Á)P1)(1Á(Á)Á:p(º =
These can be used to give the general form referred to as the class of single parameter or S-
Gini indices:
+= 1
0
dp)Á;p(º))p(Lp()Á(I
Here Á is set by the analyst to reflect societys ethical concern over inequality among the
poorest. Note that when Á = 2, then the result is the standard Gini index of inequality. Re-
sults for the S-Gini index are reported in Table H-1. The table shows how the index in-
9 Some of the analysis presented here was carried out using the software Distributive Analysis/Analyse Dis-
tributive (DAD) created by researchers at University of Laval and freely available at http://132.203.59.36:83
(see also Duclos and Araar 2006).
- 88 -
creases with the value of Á and that the increase occurs faster in urban areas indicating
greater inequality among the poorest.
The Atkinson index provides an alternative measure of inequality which explicitly incorpo-
rates normative judgments about social welfare. The index is based on an additive social
welfare function and can be expressed as:
=
+
`+
=
1µwhere,
µ
)dp)p(Qlnexp(
1
1µwhere,
µ
)dp)p(Q(
1
)µ(I
1
0
)µ1
1
1
0
)µ1(
The weighting parameter µ reflects societys aversion to inequality. By specifying different
values of this parameter one can vary the importance society attaches to mean living stan-
dards versus equality. If µ = 0, an increase in the income of a poor individual or household
has the same effect on social welfare as an increase in the income of a rich individual by
the same amount. When µ > 0, more weight is given to inequality at the lower end of the
distribution and thus an increase in the income of the poor becomes more socially desir-
able. When µ = , then society is concerned only with the poorest household. In Table H-1,
results for different values of µ are computed. Again the results suggest greater levels of
inequality among the poor in urban areas.
Table H-1: Different measures of inequality in Nambia
Urban Rural Total
S-Gini (Á = 1.5) 0.41 0.45 0.47
S-Gini (Á = 2) 0.58 0.58 0.63
S-Gini (Á = 2.5) 0.67 0.64 0.70
Atkinson (µ = 0.5) 0.27 0.30 0.32
Atkinson (µ = 1) 0.47 0.45 0.51
Atkinson (µ = 2) 0.69 0.60 0.69
Theil entropy index (¸ = 1) 0.61 0.86 0.81
Mean log deviation (¸ = 0) 0.63 0.60 0.72
Coefficient of variation 1.38 2.44 1.86
Quantile ratio (0.25;0.75) 0.22 0.38 0.26
Mean (N$) 1705.76 659.14 1083.03
- 89 -
The final set of inequality measures considered here are those of the Generalised entropy
class.
( )
1=+
0=+
0,1`+1
=
¸if,dp
µ
)p(Q
ln)
µ
)p(Q
(
¸if,dp)
)p(Q
µ
ln(
¸if,1dp)
µ
)p(Q
(
)1¸(¸
)¸(I
1
0
1
0
1
0
¸
When ¸ = 0, the index yields the mean log deviation or Theil L index reported in Table H-
1 and when ¸ = 1, the index is the Theil T measure of inequality. When ¸ = 0, then the
within group inequality contributions do not depend on the mean income of the groups and
the inequality measure is strictly population weighted.
Decomposing inequality in Namibia
One common application of the generalised entropy class of inequality indices is to de-
compose it into the contributions to overall inequality from inequality between and within
different population groups. Table H-2 shows the results from such a decomposition of the
Theil entropy index (setting ¸ = 1) index by the different economic and social groups in-
cluded in the poverty profile above. The results reveal that inequality in Namibia is a prod-
uct not so much of differences between the various population sub-groups as it is of differ-
ences within the same sub-group. For instance, gender-related inequality can almost en-
tirely (97.61 percent) be attributed to inequalities within male and female-headed house-
holds separately and much less (2.39 percent) between the two gender sub-groups. It is also
interesting to note that overall inequality is driven more by inequalities within the regions
and less so between them. This suggests that intra-regional transfers would be even more
important in addressing inequality than inter-regional transfers. The two sub-groups where
between-group inequality is highest are education and language.
This is an indication that a large part of the inequality that exists in Namibia is a result of
differences in education levels and differences between language or language/ethnic
groups. It is particularly worth noting that by organising the population according to just
four educational partitions it is possible to explain almost half of the total inequality in
Namibia. One hypothesis that can help explain why education is such an important deter-
minant of inequality is the high returns to education associated with the opening of the la-
bour market after Independence. The high level of between-group inequality among the
language group partition is a reflection of the lingering effects of the practice of discrimina-
tion in access to social and economic opportunities prior to 1990. The results remain
broadly unchanged when the decomposition is conducted on the mean log deviation (¸ = 0)
as presented in Table H-3.
- 90 -
Table H-2: Group decomposition of the Theil entropy index (¸ = 1), 2003/2004
Group
Number of
categories
in group
Within-
group
Between-
group Total
Within-
group
Between-
group Total
Percentage share
Gender 2 0.79 0.02 0.81 97.61 2.39 100.00
Age 12 0.78 0.03 0.81 95.84 4.16 100.00
Locality 2 0.70 0.11 0.81 86.43 13.57 100.00
Region 13 0.63 0.18 0.81 77.60 22.40 100.00
Language 11 0.49 0.31 0.81 61.11 38.89 100.00
Education 4 0.45 0.36 0.81 55.73 44.27 100.00
Main source of
income
14 0.63 0.18 0.81 77.69 22.31 100.00
Table H-3: Group decomposition of the Mean log deviation (¸ = 0), 2003/2004
Group
Number of
categories
in group
Within-
group
Between-
group Total
Within-
group
Between-
group Total
Percentage share
Gender 2 0.70 0.02 0.72 97.21 2.79 100.00
Age 12 0.68 0.04 0.72 94.74 5.26 100.00
Locality 2 0.61 0.11 0.72 84.51 15.49 100.00
Region 13 0.54 0.18 0.72 74.65 25.35 100.00
Language 11 0.47 0.25 0.72 65.20 34.80 100.00
Education 4 0.39 0.33 0.72 54.15 45.85 100.00
Main source of
income
14 0.53 0.19 0.72 73.86 26.14 100.00
Polarisation
The conventional inequality measures such as the Lorenz curve and the Gini-coefficient
may not be able to register important changes in the income distribution. For instance, the
Gini index may not capture changes in the share of income held by the middle stratum or
more generally reflect the concentration of incomes around distinct population groups. The
concept and measures of polarisation seek to address this. Two polarisation indices are cal-
culated for Namibia. The first is the Wolfson measure, which assumes two groups of equal
size and like the Gini index, it is between 0 (no polarization) and 1 (complete polarization).
Following Wolfson (1992) this polarisation index is given by:
tanm
2/GiniT
P =
Where T = 0.5-, L(0.5) represents the difference between 50% and the income share of the
bottom half of the population and mtan = median/mean. In the hypothetical situation of
- 91 -
perfect equality, there is also zero polarisation. However, while perfect inequality implies
that one person has all of the income, maximum polarization occurs when half the popula-
tion has zero income and the other half has twice the mean.
The second polarisation measure computed for the report is the Duclos-Esteban-Ray (DER)
index, which allows for individuals not to be clustered around discrete income intervals and
avoids arbitrary choices in the number of income groups through the use of non-parametric
estimation techniques (Duclos et al 2004). Table H-4 presents the results from the analysis
on the NHIES data regarding polarization in Namibia, by locality and in each of the 13
administrative regions. The Wolfson and DER indices for Namibia are 0.697 and 0.369,
respectively. For both indices, the values are higher in urban areas than in rural areas indi-
cating that polarization is greater in urban areas. While global data on polarisation is in-
complete, in a recent analysis researchers in Argentina computed and compared the DER
index for 35 countries in Europe, Latin America and elsewhere, and find values ranging
from 0.15 to 0.35.10 What these results suggest in other words is that not only is Namibia
one of the most unequal societies in the world when it comes to income distribution, it also
appears to be among the most polarised. Measures of polarisation as well as a broader
range of inequality indicators as presented above could be added to the indicators in the
national poverty monitoring system to track developments over time.
Table H-4: Measures of polarisation, 2003/2004
Foster-Wolfson
index
Duclos, Esteban and
Ray index*
Namibia 0.697 0.369
Urban 0.690 0.337
Rural 0.430 0.335
Caprivi 0.400 0.279
Erongo 0.678 0.347
Hardap 0.746 0.398
Karas 0.727 0.365
Kavango 0.481 0.323
Khomas 0.762 0.346
Kunene 0.443 0.298
Ohangwena 0.332 0.284
Omaheke 0.711 0.382
Omusati 0.322 0.277
Oshana 0.527 0.332
Oshikoto 0.414 0.311
Otjozondjupa 0.624 0.351
* ±=0.5
10 See: Income Polarisation: An exploratory analysis for Latin America by Leonardo Gasparini, Matías
Horenstein, Ezequiel Molina and Sergio Olivieri, unpublished working paper at Universidad Nacional de La
Plata (Argentina).
- 92 -
ANNEX I: Confidence intervals
Table I-1: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by region
Estimate Std.
Err.
[95%
Conf.
Interval] Deff
Severely poor (p0_185)
Caprivi 0.125 0.018 0.091 0.160 2.043
Erongo 0.048 0.014 0.021 0.076 2.257
Hardap 0.219 0.037 0.146 0.291 5.098
Karas 0.125 0.027 0.073 0.178 4.178
Kavango 0.367 0.035 0.299 0.436 3.764
Khomas 0.024 0.004 0.015 0.032 0.996
Kunene 0.131 0.046 0.040 0.222 9.020
Ohangwena 0.193 0.035 0.124 0.262 5.687
Omaheke 0.175 0.036 0.104 0.246 4.497
Omusati 0.128 0.024 0.081 0.176 5.048
Oshana 0.078 0.013 0.052 0.103 2.377
Oshikoto 0.166 0.019 0.128 0.204 2.697
Otjozondjupa 0.158 0.021 0.117 0.199 2.296
Urban 0.060 0.006 0.049 0.072 2.639
Rural 0.191 0.011 0.170 0.212 3.992
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Caprivi 0.286 0.028 0.232 0.340 2.716
Erongo 0.103 0.027 0.050 0.156 4.242
Hardap 0.321 0.042 0.238 0.403 5.219
Karas 0.219 0.032 0.157 0.281 3.762
Kavango 0.565 0.031 0.503 0.626 2.825
Khomas 0.063 0.008 0.047 0.078 1.194
Kunene 0.230 0.057 0.117 0.342 8.796
Ohangwena 0.447 0.045 0.359 0.536 5.927
Omaheke 0.301 0.050 0.203 0.399 5.840
Omusati 0.310 0.029 0.253 0.368 3.854
Oshana 0.196 0.023 0.152 0.241 3.258
Oshikoto 0.408 0.027 0.356 0.460 2.916
Otjozondjupa 0.278 0.027 0.224 0.331 2.570
Urban 0.120 0.009 0.103 0.137 3.111
Rural 0.382 0.013 0.356 0.407 3.926
Namibia 0.276 0.009 0.259 0.293 3.807
- 93 -
Table I-2: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by region
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
San 0.390 0.063 0.266 0.514 3.021
Caprivi languages 0.108 0.016 0.077 0.139 1.925
Otjiherero 0.088 0.020 0.048 0.128 4.012
Rukavango 0.349 0.032 0.286 0.411 3.697
Nama/Damara 0.214 0.017 0.181 0.247 2.524
Oshiwambo 0.118 0.010 0.097 0.138 4.420
Setswana 0.010 0.011 -0.012 0.032 0.593
Afrikaans 0.035 0.009 0.017 0.053 2.741
German 0.000 0.000 0.000 0.000 .
English 0.004 0.004 -0.004 0.011 0.518
Others 0.096 0.033 0.031 0.161 1.306
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
San 0.597 0.048 0.504 0.691 1.689
Caprivi languages 0.246 0.025 0.197 0.295 2.468
Otjiherero 0.170 0.029 0.112 0.228 4.760
Rukavango 0.544 0.028 0.490 0.599 2.554
Nama/Damara 0.342 0.022 0.299 0.386 3.184
Oshiwambo 0.285 0.014 0.258 0.313 4.164
Setswana 0.145 0.045 0.056 0.234 0.793
Afrikaans 0.079 0.011 0.057 0.101 1.925
German 0.000 0.000 0.000 0.000 .
English 0.006 0.004 -0.003 0.014 0.437
Others 0.164 0.047 0.072 0.257 1.659
Namibia 0.276 0.009 0.259 0.293 3.807
- 94 -
Table I-3: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by sex of the head of household
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Female 0.151 0.009 0.134 0.168 2.271
Male 0.129 0.008 0.114 0.144 3.096
Not stated 0.124 0.062 0.003 0.245 1.193
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Female 0.304 0.012 0.280 0.328 2.694
Male 0.258 0.009 0.239 0.276 2.734
Not stated 0.150 0.065 0.022 0.278 1.132
Namibia 0.276 0.009 0.259 0.293 3.807
Table I-4: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by type of dwelling
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Detached house 0.035 0.004 0.027 0.043 1.604
Apartment 0.013 0.006 0.001 0.026 0.842
Traditional dwelling 0.213 0.013 0.187 0.239 3.793
Improvised housing 0.185 0.015 0.155 0.214 2.749
Other 0.082 0.022 0.038 0.125 1.928
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Detached house 0.081 0.006 0.069 0.093 1.856
Apartment 0.027 0.009 0.008 0.045 0.992
Traditional dwelling 0.436 0.015 0.406 0.466 3.473
Improvised housing 0.319 0.019 0.283 0.356 3.019
Other 0.148 0.031 0.088 0.209 2.186
Namibia 0.276 0.009 0.259 0.293 3.807
- 95 -
Table I-5: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by main water source
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Piped in dwelling 0.016 0.003 0.011 0.021 1.157
Piped on site 0.099 0.010 0.079 0.119 1.823
Neighbour's tap 0.192 0.021 0.150 0.234 1.794
Public tap 0.190 0.015 0.161 0.219 3.539
Water carrier/tanker 0.232 0.039 0.155 0.309 3.243
Private borehole 0.353 0.086 0.184 0.522 0.553
Communal borehole 0.166 0.043 0.082 0.250 1.071
Protected well 0.228 0.029 0.170 0.286 3.244
Spring 0.343 0.042 0.260 0.426 2.877
Flowing water 0.175 0.031 0.114 0.235 1.005
Rain water tank 0.203 0.046 0.114 0.293 2.489
Unprotected well 0.191 0.030 0.132 0.249 2.226
Dam/pool/stagnant water 0.109 0.105 -0.098 0.315 1.365
Other 0.173 0.051 0.072 0.274 0.797
Not stated 0.211 0.148 -0.079 0.502 0.524
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262) 0.040 0.005 0.031 0.049 1.511
Piped in dwelling 0.216 0.017 0.183 0.249 2.647
Piped on site 0.382 0.027 0.328 0.436 1.923
Neighbour's tap 0.390 0.018 0.354 0.426 3.437
Public tap 0.449 0.044 0.363 0.535 2.902
Water carrier/tanker 0.621 0.100 0.424 0.818 0.725
Private borehole 0.260 0.052 0.158 0.362 1.146
Communal borehole 0.434 0.035 0.365 0.503 3.351
Protected well 0.519 0.044 0.434 0.605 2.775
Spring 0.409 0.044 0.323 0.495 1.194
Flowing water 0.435 0.050 0.337 0.533 1.978
Rain water tank 0.397 0.039 0.321 0.473 2.390
Unprotected well 0.201 0.133 -0.060 0.463 1.322
Dam/pool/stagnant water 0.309 0.070 0.172 0.446 0.979
Other 0.492 0.290 -0.076 1.061 1.341
Not stated 0.276 0.009 0.259 0.293 3.807
Namibia 0.016 0.003 0.011 0.021 1.157
- 96 -
Table I-6: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by main toilet facility
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Flush/sewer 0.023 0.004 0.016 0.030 1.834
Flush/septic tank 0.038 0.012 0.015 0.060 1.314
Pit latrine/VIP 0.081 0.014 0.053 0.109 1.207
Pit latrine/no ventilation 0.166 0.027 0.112 0.219 2.869
Bucket 0.248 0.058 0.134 0.362 2.860
Other 0.102 0.070 -0.035 0.239 1.003
Bush 0.216 0.011 0.194 0.239 3.798
Not stated 0.039 0.039 -0.038 0.116 0.651
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Flush/sewer 0.059 0.006 0.047 0.071 2.198
Flush/septic tank 0.103 0.020 0.064 0.142 1.496
Pit latrine/VIP 0.168 0.024 0.121 0.214 1.745
Pit latrine/no ventilation 0.336 0.031 0.275 0.397 2.324
Bucket 0.404 0.061 0.284 0.525 2.483
Other 0.222 0.097 0.032 0.412 1.029
Bush 0.423 0.013 0.397 0.450 3.615
Not stated 0.169 0.120 -0.066 0.403 1.631
Namibia 0.276 0.009 0.259 0.293 3.807
Table I-7: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by material of wall of housing
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Cement blocks 0.037 0.004 0.029 0.044 1.791
Bricks 0.072 0.017 0.039 0.105 1.016
Iron/zinc 0.198 0.015 0.169 0.226 2.371
Poles/sticks/grass 0.189 0.020 0.150 0.228 3.001
Sticks/mud/clay/dung 0.231 0.016 0.199 0.262 2.989
Asbetos 0.054 0.025 0.004 0.104 0.966
Tiles 0.127 0.063 0.003 0.250 1.140
Slates 0.000 0.000 0.000 0.000 .
Thatch 0.361 0.075 0.213 0.508 2.604
Other 0.564 0.348 -0.119 1.247 0.492
Not stated 0.161 0.033 0.097 0.226 1.929
Dont know 0.342 0.178 -0.008 0.691 1.125
Namibia 0.138 0.007 0.125 0.152 3.891
- 97 -
Poor (p0_262)
Cement blocks 0.094 0.008 0.079 0.109 2.886
Bricks 0.166 0.028 0.112 0.221 1.335
Iron/zinc 0.333 0.017 0.300 0.366 2.287
Poles/sticks/grass 0.413 0.025 0.364 0.462 3.014
Sticks/mud/clay/dung 0.461 0.018 0.426 0.496 2.598
Asbetos 0.195 0.046 0.104 0.286 1.045
Tiles 0.127 0.063 0.003 0.250 1.140
Slates 0.174 0.178 -0.176 0.525 0.885
Thatch 0.492 0.080 0.335 0.649 2.715
Other 0.564 0.348 -0.119 1.247 0.492
Not stated 0.270 0.042 0.189 0.352 2.109
Dont know 0.342 0.178 -0.008 0.691 1.125
Namibia 0.276 0.009 0.259 0.293 3.807
Table I-8: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by material of roof of housing
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Cement blocks 0.018 0.014 -0.009 0.045 0.538
Bricks 0.075 0.062 -0.047 0.196 1.555
Iron/zinc 0.093 0.006 0.082 0.105 2.624
Poles/sticks/grass 0.192 0.026 0.140 0.244 2.966
Sticks/mud/clay/dung 0.182 0.069 0.047 0.316 1.803
Asbetos 0.013 0.007 0.000 0.027 1.411
Tiles 0.000 0.000 0.000 0.000 .
Slates 0.130 0.070 -0.008 0.268 0.920
Thatch 0.232 0.017 0.198 0.266 3.746
Other 0.538 0.352 -0.153 1.228 0.497
Not stated 0.110 0.033 0.046 0.174 1.464
Dont know 0.289 0.208 -0.120 0.699 1.268
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Cement blocks 0.063 0.033 -0.002 0.129 0.960
Bricks 0.225 0.095 0.038 0.411 1.447
Iron/zinc 0.180 0.008 0.164 0.197 2.986
Poles/sticks/grass 0.403 0.030 0.343 0.462 2.531
Sticks/mud/clay/dung 0.290 0.077 0.138 0.441 1.656
Asbetos 0.051 0.016 0.019 0.083 2.195
Tiles 0.000 0.000 0.000 0.000 .
Slates 0.171 0.082 0.105 0.331 0.986
Thatch 0.470 0.020 0.431 0.508 3.408
Other 1.000 0.000 1.000 1.000 .
Not stated 0.179 0.041 0.097 0.260 1.572
- 98 -
Dont know 0.402 0.211 -0.012 0.817 1.112
Namibia 0.276 0.009 0.259 0.293 3.807
Table I-9: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by material of floor of housing
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Sand 0.218 0.013 0.192 0.244 3.231
Concrete 0.051 0.004 0.043 0.059 1.635
Mud/clay/dung 0.226 0.020 0.187 0.265 3.136
Wood 0.013 0.011 -0.009 0.034 0.661
Other 0.099 0.047 0.008 0.191 1.502
Not stated 0.542 0.351 -0.147 1.232 0.496
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Sand 0.419 0.015 0.390 0.448 2.859
Concrete 0.113 0.007 0.100 0.126 2.303
Mud/clay/dung 0.461 0.022 0.417 0.504 2.698
Wood 0.087 0.049 -0.010 0.183 1.997
Other 0.134 0.051 0.034 0.235 1.385
Not stated 0.542 0.351 -0.147 1.232 0.496
Namibia 0.276 0.009 0.259 0.293 3.807
Table I-10: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by ownership of and access to radio
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Owns 0.114 0.007 0.101 0.127 3.011
Does not own, has access 0.206 0.016 0.175 0.236 1.929
Neither own nor has ac-
cess
0.193 0.014 0.164 0.221 1.983
Not stated 0.136 0.097 -0.055 0.327 1.130
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Owns 0.236 0.009 0.219 0.254 3.040
Does not own, has access 0.399 0.019 0.362 0.435 1.870
Neither own nor has ac-
cess
0.355 0.019 0.318 0.391 2.208
Not stated 0.308 0.134 0.045 0.571 1.181
Namibia 0.276 0.009 0.259 0.293 3.807
- 99 -
Table I-11: 95 % confidence intervals for estimates of incidence of poor and severely
poor households by ownership of and access to telephone
Estimate Std. Err. [95%
Conf.
Interval] Deff
Severely poor (p0_185)
Owns 0.019 0.003 0.013 0.024 1.319
Does not own, has access 0.155 0.010 0.136 0.175 2.497
Neither own nor has access 0.241 0.014 0.213 0.270 3.505
Not stated 0.200 0.094 0.015 0.385 1.329
Namibia 0.138 0.007 0.125 0.152 3.891
Poor (p0_262)
Owns 0.049 0.005 0.040 0.057 1.455
Does not own, has access 0.335 0.013 0.310 0.360 2.469
Neither own nor has access 0.446 0.017 0.413 0.479 3.525
Not stated 0.413 0.107 0.202 0.624 1.140
Namibia 0.276 0.009 0.259 0.293 3.807
Republic of Namibia
CENTRAL BUREAU OF STATISTICS
NATIONAL PLANNING COMMISION
ISBN 978 - 0 - 86976 - 781 - 8