Phase II – Selected Studies

Assessing the Macroeconomic Impact
of HIV/AIDS in Uganda

Phase II Report: Selected Studies

1. The impact of HIV/AIDS on poverty
2. Assessing sectoral vulnerability to HIV/AIDS
3. HIV costing, financing and expenditure
4. The demographic impact of HIV/AIDS
5. Macroeconomic linkages between aid flows, the
exchange rate, inflation and exports

Keith Jefferis, Sebastian Baine, Jimrex Byamugisha & Justine
Nannyonjo

Draft Final Report

Submitted to:

Ministry of Finance Planning and Economic Development
United Nations Development Programme
Kampala
2008

Contents
Abbreviations ........................................................................................................................................... i
Chapter 1: Evaluating the Macroeconomic Impact of HIV/AIDS in Uganda: Phase 2 –
Selected Studies
1. Introduction .................................................................................................................................... 1
2. Summary of Results of Mini-studies ............................................................................................... 2
Modelling of Household and Poverty Impact ..................................................................................... 2
Summary of Results ........................................................................................................................ 2
Modelling of Sectoral HIV-vulnerability/risk Exposure ....................................................................... 3
HIV Costing, Financing and Expenditure ............................................................................................. 4
Demographic Projections.................................................................................................................... 5
Analytical (Econometric) Studies on Macroeconomic Relationships ................................................. 6
Chapter 2: The Impact of HIV/AIDS on Poverty
1. Introduction .................................................................................................................................... 8
2. Background: HIV/AIDS and Poverty in Uganda ............................................................................... 8
HIV/AIDS in Uganda ............................................................................................................................ 8
Provision of ART ................................................................................................................................ 10
Poverty Levels ................................................................................................................................... 12
HIV/AIDS and Poverty ....................................................................................................................... 13
Impact of HIV/AIDS ....................................................................................................................... 13
Household Level: HIV/AIDS, Poverty and Inequality ........................................................................ 14
Long-term Household Costs of HIV/AIDS.......................................................................................... 15
Household Coping Strategies ............................................................................................................ 15
3. Modelling the Household Level Impact of HIV/AIDS .................................................................... 16
Methodology and Previous Studies .................................................................................................. 16
Approach followed in this Study ....................................................................................................... 17
Simulating HIV/AIDS...................................................................................................................... 17
Indicators ...................................................................................................................................... 19
Poverty Data ..................................................................................................................................... 20
4. Results ........................................................................................................................................... 21
Short-term Impact ............................................................................................................................ 21
Impact of Health and Related Costs.............................................................................................. 21
Impact of Funeral Costs ................................................................................................................ 22
Income Adjustment Effect ............................................................................................................ 22
Income and Expenditure Effects Combined.................................................................................. 23
Long-term HIV/AIDS impact .............................................................................................................. 24
HIV/AIDS Impact on Poverty Gap and Severity Indices .................................................................... 24
Further Understanding of Health Costs ............................................................................................ 27
5. Conclusions ................................................................................................................................... 28

Chapter 3: Assessing Sectoral Vulnerability to HIV/AIDS
1. Introduction .................................................................................................................................. 29
2. Sectoral Impact – HIV Prevalence ................................................................................................. 30
3. Sectoral Impact – Occupational Structure .................................................................................... 35
4. Cost of Providing Anti-Retroviral Therapy (ART) .......................................................................... 41
5. Summary and Conclusion ............................................................................................................. 42
6. Appendix: Data Tables .................................................................................................................. 44
Chapter 4: HIV Costing, Financing and Expenditure
1. Introduction .................................................................................................................................. 45
2. Methodology................................................................................................................................. 46
3. Results ........................................................................................................................................... 47
4. Future Resource Needs and Funding. ........................................................................................... 50
5. Conclusion ..................................................................................................................................... 51
Appendix 1 – Data Collection Form ...................................................................................................... 52
Appendix 2: PEPFAR Uganda Partners: FY 2006 ................................................................................... 53
Appendix 3: Responses Received.......................................................................................................... 54
Chapter 5: The Demographic Impact of HIV/AIDS
1. Introduction .................................................................................................................................. 55
2. Scenarios ....................................................................................................................................... 55
3. Population Projections .................................................................................................................. 56
Basic Demographic Assumptions ...................................................................................................... 56
HIV/AIDS Projections ........................................................................................................................ 59
Treatment Scenarios ..................................................................................................................... 61
4. Population Estimates .................................................................................................................... 62
5. Further Analysis of the Impact of HIV/AIDS and the Provision of ART ......................................... 66
6. Conclusions and Implications........................................................................................................ 73
Spectrum Policy Modelling System....................................................................................................... 75
DemProj ................................................................................................................................................ 76
AIDS Impact Module (AIM) ................................................................................................................... 80
Chapter 6: An Empirical Analysis of the Link between Aid Flows, the Exchange Rate,
Inflation in Uganda
1. Introduction .................................................................................................................................. 85
2. Literature Survey........................................................................................................................... 87
Aid flows, Exchange rate and Inflation ............................................................................................. 87
Real Exchange Rate and Exports ....................................................................................................... 90
3. Descriptive Analysis ...................................................................................................................... 95
Liquidity Injection and Money Growth. ............................................................................................ 96
Sterilisation and its Costs ................................................................................................................. 97
Broad Money (M3) Growth and Inflation in Uganda ........................................................................ 98
Inflation and Real Effective Exchange Rates (REER) ......................................................................... 98
Real Effective Exchange Rate (REER) and Export Receipts ............................................................... 99
4. Estimation Methods and Data .................................................................................................... 105
5. Estimation Results ....................................................................................................................... 106

Identifying the Effects of Aid........................................................................................................... 106
Real Exchange Rate and Exports ..................................................................................................... 107
6. Conclusions and Policy Recommendations................................................................................. 108
Appendices.......................................................................................................................................... 111
Appendix 1: Models and Estimation Methods................................................................................ 111
Aid Flows, Inflation and the Real Exchange Rate ........................................................................ 111
Export Supply Equation ............................................................................................................... 112
Estimating Exchange Rate Volatility............................................................................................ 113
Appendix 2: Results......................................................................................................................... 115
Stationarity Test Results – Export Equation................................................................................ 115
Cointegration Analysis ................................................................................................................ 116
Error Correction Model ............................................................................................................... 119
Identifying the Effects of Aid....................................................................................................... 123
Appendix 3: Computation of the REER ........................................................................................... 127
References .......................................................................................................................................... 129

List of Figures
Figure 1: HIV Prevalence Rates, 1988-2005 ............................................................................................ 9
Figure 2: Adult HIV Prevalence Rates by Region................................................................................... 10
Figure 3: Total Number of People on ART ............................................................................................ 11
Figure 4: Trends in Poverty and Inequality ........................................................................................... 12
Figure 5: Regional Poverty Rates .......................................................................................................... 13
Figure 6: Change in P2 Values as a Result of HIV/AIDS ......................................................................... 26
Figure 7: Effects of Different Scenarios on Poverty Gap....................................................................... 26
Figure 8: Poverty Rates and the Severity of Poverty ............................................................................ 27
Figure 9: Contribution of Health Costs to Total Household Consumption for HIV and No-HIV Scenarios
(%) ......................................................................................................................................................... 27
Figure 10: HIV Prevalence by Residence, Education, Work Status and Wealth ................................... 31
Figure 11: HIV Prevalence by Sector ..................................................................................................... 32
Figure 12: Distribution of HIV+ Workers by Sector ............................................................................... 33
Figure 13: HIV Prevalence by Gender and Sector ................................................................................. 34
Figure 14: Distribution of HIV+ Workers by Sector and Gender ........................................................... 35
Figure 15: HIV Prevalence by Occupation ............................................................................................. 35
Figure 16: “Proxy” HIV Prevalence Rates by Sector (derived from occupations) ................................. 36
Figure 17: Index of Sectoral Vulnerability ............................................................................................. 37
Figure 18: Total Cost of Educating HIV+ Workers (% by sector) ........................................................... 39
Figure 19: Cost of Education of Average Worker, by Sector (Shs million) ............................................ 40
Figure 20: Cost of Replacing HIV+ Workers (as percentage of annual wage bill) ................................. 41
Figure 21: Average Wage by Sector (US$/year).................................................................................... 42
Figure 22: HIV/AIDS Funding and Expenditure ..................................................................................... 45
Figure 23: Financing of HIV/AIDS-related Spending, 2003/4 - 2006/7 ................................................. 46
Figure 24: Breakdown of HIV/AIDS Spending 2004/5 - 2006/7 ............................................................ 49
Figure 25: Breakdown of PEPFAR Spending, 2005-7 ............................................................................ 49
Figure 26: International Migration........................................................................................................ 58

Figure 27: Life Expectancy, Male & Female .......................................................................................... 59
Figure 28: Adult HIV Prevalence ........................................................................................................... 60
Figure 29: ART Rollout .......................................................................................................................... 62
Figure 30: Total Population ................................................................................................................... 63
Figure 31: Population Deficit due to HIV/AIDS ..................................................................................... 64
Figure 32: Population Growth Rates ..................................................................................................... 65
Figure 33: Number of People Infected with HIV/AIDS.......................................................................... 67
Figure 34: Adult HIV Prevalence ........................................................................................................... 67
Figure 35: No. of Adults Receiving ART ................................................................................................. 68
Figure 36: AIDS-related Deaths ............................................................................................................. 69
Figure 37: Life Expectancy .................................................................................................................... 70
Figure 38: Age-specific AIDS-related Deaths in 2015............................................................................ 71
Figure 39: Projected Loss of Labour Force (15-59 yrs) ......................................................................... 72
Figure 40: Orphan Projections (Double & Single) ................................................................................. 73
Figure 41: Nominal and Real Exchange Rates ....................................................................................... 87
Figure 42: External Aid Transfers (Grants) to Uganda, 1995/6 - 2006/7 .............................................. 96
Figure 43: Liquidity Injection and Money Supply Growth .................................................................... 97
Figure 44: Inflation and Broad Money (M3) Growth in Uganda .......................................................... 98
Figure 45: Nominal and Real Effective Exchange Rates and Inflation (annual averages) ..................... 99
Figure 46: Evolution of the REER and Total Export Receipts .............................................................. 101
Figure 47: Total Exports (US$ million) and Terms of Trade ................................................................ 101
Figure 48: Coffee Exports (volume) and REER .................................................................................... 102
Figure 49: Tea Exports (volume) and REER ......................................................................................... 103
Figure 50: Cotton Exports (volume) and REER .................................................................................... 103
Figure 51: Fish Exports (volume) and REER ........................................................................................ 104
Figure 52: Maize Exports (volume) and REER ..................................................................................... 104
Figure 53: Flower Exports (value, US$ million) and REER ................................................................... 105
Figure 54: Identifying the Effects of aid using Headline Inflation ....................................................... 109
Figure 55: Identifying the Effects of aid using Underlying Inflation ................................................... 110

List of Tables
Table 1: Impact of HIV/AIDS on Poverty and Inequality in Selected Countries .................................... 17
Table 2: Adult Equivalent Conversion Factors ...................................................................................... 20
Table 3: Adult-equivalent PDL Estimates as used in 2005/2006 UNHS ................................................ 21
Table 4: Poverty Rates by Region.......................................................................................................... 21
Table 5: Changes in Poverty Level: Health Costs Impact ...................................................................... 22
Table 6: Changes in Poverty Level: Funeral Costs Impact..................................................................... 22
Table 7: Changes in Poverty Level: Income Adjustment Effect ............................................................ 23
Table 8: Poverty Levels Incorporating Income Adjustments, Health and Funeral Costs Effect............ 23
Table 9: Changes in Poverty Levels (percentage points and %)............................................................ 24
Table 10: Long-term changes in Poverty Level: Income Adjustments due to Death of HIV+ Income
Earners .................................................................................................................................................. 24
Table 11: The Effect of Different Scenarios on the Poverty Gap (P1) and Severity (P2) ...................... 25
Table 12: Median Monthly Wages by Occupation................................................................................ 37

Table 13: Education Level by Sector (% of workforce in sector)........................................................... 38
Table 14: Cost of Education .................................................................................................................. 38
Table 15: HIV/AIDS-related Spending by Institution, 2004/05 – 2006/07 (US$) .................................. 48
Table 16: External Component of HIV/AIDS-related Spending (2004/05 – 2006/07, US$) .................. 50
Table 17: External Component of PEPFAR Spending (FY2005-2007, US$) ........................................... 50
Table 18: Population 1980 by Age & Gender........................................................................................ 57
Table 19: Adult HIV Prevalence, 2004/05 (%) ....................................................................................... 60
Table 20: Ratio of Fertility of HIV Infected Women to the Total Fertility of Uninfected Women ........ 61
Table 21: Cumulative Percentage Dying from HIV/AIDS by Number of Years since Infection, without
ART ........................................................................................................................................................ 61
Table 22: Population Projections by Age and Gender for 2025 (million) ............................................. 65
Table 23: Population Projections by Age and Gender for 2025 (million) – “with ART” Scenarios ....... 66
Table 24: Government Securities and Interest costs (billion shillings) (as at June).............................. 97
Table 25: Selected Exports Receipts (million US$).............................................................................. 100
Table 26: Share of Total Exports ......................................................................................................... 100
Table 27: Responsiveness of Exports to 1% Change in Independent Variable ................................... 108

Abbreviations
AIDS Acquired Immune Deficiency Syndrome
ART Anti-retroviral therapy
BoU Bank of Uganda
CGE Computable General Equilibrium
DTIS Diagnostic Trade Integration Study
ETOs Extra-territorial Organisations
GDP Gross Domestic Product
GFATM Global Fund for AIDS, Tuberculosis and Malaria
GOU Government of Uganda
HIPC Heavily Indebted Poor Countries
HIV Human Immunodeficiency Virus
IEC Information, Education and Communication
IMF International Monetary Fund
MDGs Millennium Development Goals
MoFPED Ministry of Finance, Planning and Economic Development
MOH Ministry of Health
MTCT Mother to Child Transmission
MTEF Medium Term Expenditure Framework
NACP National AIDS Control Programme
NGOs Non Governmental Organisations
NSF National Strategic Framework
NSP National Strategic Plan
ODA Official Development Assistance
PEAP Poverty Eradication Action Plan
PEPFAR President’s Emergency Plan for AIDS Relief
PMTCT Prevention of Mother-To-Child Transmission of HIV
PRSPs Poverty Reduction Strategy Papers -
REER Real Effective Exchange Rate
STD Sexual Transmitted Diseases
TASO The AIDS Support Organisation
UAC Uganda AIDS Commission
UN United Nations
UNAIDS Joint United Nations Programme on HIV/AIDS
UNASO Uganda Network of AIDS Service Organisations
UNDP United Nations Development Programme
UNICEF United Nations Children’s Fund
USAID United States Aid for International Development
US$ United States Dollar
WHO World Health Organisation

Chapter 1: Evaluating the
Macroeconomic Impact of
HIV/AIDS in Uganda: Phase
2 – Selected Studies
1. Introduction
This report is the second in a series of reports on Evaluating the Macroeconomic Impact of HIV/AIDS
in Uganda, commissioned by the Ministry of Finance, Planning and Economic Development
(MoFPED) and the United Nations Development Programme (UNDP) in Uganda. Although there is
awareness of the general economic impacts of HIV and AIDS in Uganda, little has been done to
quantify these impacts, particularly the impact of alternative policies in dealing with the scourge.
The lack of quantitative information has hindered macroeconomic planning and the formulation of
an appropriate HIV/AIDS response. In particular, while there is awareness of the need to scale-up
the response to HIV and AIDS, there is concern that the macroeconomic stability which Uganda has
made over the past 15 years, could be lost. The result has been uncertainty over the level of
investment that should be made in responding to HIV/AIDS in Uganda. There has also been a lack of
understanding as to whether the benefits of a rapid scale-up of treatment would be primarily
economic, social, or both. This study endeavours to fill that gap.
This study was conducted in three phases, as follows:
Phase I: Literature review from Uganda and the African region on existing micro economic and
macroeconomic studies and models, detailed methodology and scope of work for phase
two;
Phase II: A selected number of micro-economic studies/surveys;
Phase III: An aggregated macro-economic analysis, production and validation of report.
The draft Phase I Report, Literature Review: The Macroeconomic Impact of HIV/AIDS, was presented
to a stakeholders workshop in August, 2007. Following comments received at the workshop, the
report was revised and published in October 2007.
Phase II of the project involved conducting five mini-studies which provided essential inputs into the
macroeconomic analysis work for Phase III. The mini-studies were as follows:
1. Modelling the household and poverty impact of HIV/AIDS.
2. Modelling of sectoral HIV-vulnerability/risk exposure.
3. HIV costing, financing and expenditure.
4. Preparation of demographic projections.
5. Analytical (econometric) studies on macroeconomic relationships between aid flows, inflation,
exchange rates and exports.
This report encompasses the results of the above mini-studies, with each study covering one
chapter of the report. This chapter gives a summary of the rationale and approaches used in each of
the studies, and the accompanying results.

1

Phase III of the study consists of a modelling exercise to quantify the impact of HIV/AIDS on the
Uganda economy, and of the impact of interventions related to the provision of Anti-retroviral
Therapy (ART). The report for Phase III, which will be produced separately, contains the modelling
results, conclusions regarding macroeconomic impact of HIV/AIDS in Uganda, and policy
recommendations.

2. Summary of Results of Mini-studies
Modelling of Household and Poverty Impact
The aim of this study was to model the impact of HIV/AIDS on household incomes and poverty
levels. It utilises techniques used previously in studies of Swaziland, Zambia, Ghana, Kenya and
Botswana, and is based on household level data (from nationwide household surveys) on incomes,
expenditure and poverty status, and data on sero-prevalence that can be matched to the household
data. The basic approach is to model the impact of HIV/AIDS on household income and expenditure,
using information on sero-prevalence for different demographic categories to determine which
individuals and households are HIV+. Poverty levels are then recalculated taking account of the
simulated impact of HIV/AIDS.

Summary of Results
The study finds that HIV/AIDS is likely to have a negative impact on poverty in the short-term.
However, the impact is likely to be fairly small, raising the overall headcount poverty rate by only 1.4
percentage points (or by 5.2%). In absolute terms, the impact is greater in rural areas, where poverty
rises by 1.6 percentage points, compared to urban areas (0.9 percentage points). However, because
poverty rates are so much lower in urban areas, the proportionate impact on urban poverty is
greater. The short-term poverty impact analysis models the effect of higher household expenditure
on healthcare and funeral costs, as well as income losses due to the ill-health of breadwinners. Of
these, the greatest impact comes from additional health-care expenditure. Estimates of the
additional expenditure burden are drawn from international studies, and further research would be
necessary to obtain Uganda-specific estimates of this impact.
The impact on poverty is also greater in regions with relatively low poverty rates. This is because in
high-poverty regions, the majority of households are already poor, so HIV/AIDS pushes fewer
households below the poverty line. When looking at the depth of poverty, however, (the degree to
which households fall below the poverty line), HIV/AIDS has a greater impact in the rural areas as it
pushes already poor households even further into poverty.
The estimated impact on poverty levels in Uganda is comparable with that found in other studies for
Kenya, which has a similar HIV prevalence rate, but is much smaller than the impact in high-
prevalence countries in Southern Africa.
The analysis also considered the long-term impact on poverty levels, which focuses on changed
income levels (due to the death of breadwinners) and changed household sizes. These two effects
counteract each other. There are also no health cost effects. Hence the long-term impact on poverty
is smaller than the short-term impact.
While this analysis does not address the impact on poverty of Anti-retroviral therapy (ART) provision,
the fact that additional health care costs are the main contributor to increased poverty levels
indicates that ART provision would have a beneficial impact. This is because ART has a significant
positive effect on health and well-being, and will therefore reduce health-related expenditure and

2

increase income levels. However, this will be offset to the extent that ART provision requires regular
visits to health facilities, which has implications for both household expenditure and time available
for work. Further research evaluating the level of health expenditure in households with HIV+
members and the impact of ART provision would therefore be worthwhile.

Modelling of Sectoral HIV-Vulnerability/Risk Exposure
This analysis focused on the extent to which different sectors of the economy are vulnerable to the
negative impacts of HIV prevalence amongst their workforces. It made use of information regarding
the variation in HIV prevalence amongst demographic and occupational groups, and of the differing
occupational and demographic structures in the workforce in differing industries.
Sectoral vulnerability was assessed in terms of various indicators:
- Reported sectoral HIV prevalence.
- Implied sectoral HIV prevalence according to occupational structures.
- The importance of skilled workers in production.
- The cost of educating workers who are lost to HIV/AIDS.

Summary of Results
With regard to sectoral HIV prevalence as determined from the results of the Sero-Prevalence
Survey, the most vulnerable sectors are Public Administration; Hotels & Restaurants; Sales and
Fishing. While the Agricultural sector has a relatively low prevalence rate, it is by far the largest
sector of the economy, and hence has the largest number of HIV positive workers in absolute terms.
The Sero-Prevalence Survey sample was too small to provide prevalence rates for the Mining and
Finance sectors.
The survey also provides information on HIV prevalence by occupation. Amongst working adults,
there is some evidence that HIV prevalence varies across occupations. The prevalence rate is
relatively high amongst Sales, Clerical and Service sector workers, who might generally be classed as
semi-skilled. There is a slightly lower, but above average, prevalence rate for skilled professional and
manual workers. The lowest prevalence rates are for the unskilled categories of manual and
agricultural workers. These results confirm that unskilled workers have lower HIV prevalence rates
than those of semi-skilled and skilled workers, which contrasts with the findings in South Africa.
The information on HIV prevalence by occupation was combined with information on the sectoral
composition of the labour force in different sectors to further analyse sectoral vulnerabilities. A
“proxy” HIV prevalence rate was calculated for each sector, combining occupational HIV prevalence
rates with the occupational structure of the various sectors. The results are largely consistent with
the earlier results, although the prominence of Public Administration is reduced. The proxy
prevalence rate for Finance suggests that it is one of the most vulnerable sectors.
Sectoral vulnerability also depends on the contribution of different groups of workers to production
and the ease with which they can be replaced should they fall sick or die. More skilled workers make
a greater contribution to production and are more difficult to replace. Hence sectoral vulnerability
depends on both prevalence rates and the proportion of skilled workers in sectoral employment.
Taking both of these into account, the Education, Health, Finance and Public Administration sectors
are the most vulnerable.

3

The study also takes into account the cost of educating workers in the different sectors. Using data
on the level of education of workers in the different sectors, and estimates of the costs of education,
as well as sectoral HIV prevalence rates, we can calculate the cost of replacing HIV+ workers as a
percentage of the annual wage bill. The greatest burden is in the Hotels & Restaurants sector, which
mainly represents the impact of relatively low wages in the sector combined with a high HIV
prevalence rate. The next highest burdens are in Education; Public Administration; Health and Social
Work, reflecting the high level of education of workers in these sectors and relatively high
prevalence rates. While the Financial sector has both a relatively high prevalence rate and highly
educated workers, it also has the highest wage levels of any sector so the relative cost of educating
workers is reduced.

HIV Costing, Financing and Expenditure
One of the major determinants of the macroeconomic impact of HIV/AIDS expenditure is the extent
to which that spending is sourced domestically (from the government budget) or externally (from
donor funds). A second important aspect is whether the funds are spent externally (on, for instance
imported drugs) or domestically. Most of the concerns about the expenditure impact of HIV/AIDS
stems from concerns that large amounts of external funds will flow into the country and boost
aggregate demand, which will in turn cause inflation and real exchange rate appreciation (and loss of
international competitiveness), and destabilise the macroeconomic achievements that have been
secured over the past 15 years. However, this effect is reduced if the greater the proportion of
spending is devoted to imported goods and services, as domestic aggregate demand will be affected
less.
While there is some information on the sourcing of HIV/AIDS funding, there is little or no
information on how the money is spent, or what it is spent on. The objective of this study was to
track the flow of resources received through to spending, to determine what HIV/AIDS related funds
are spent on, and in particular, whether that expenditure is on domestic or imported goods and
services.
The methodology followed was to gather information from resource providers (donors and
Governement of Uganda [GoU]) relating to the sources of funds, and from entities involved in
spending those resources. Information was sought on the main categories of expenditure, and on
whether that expenditure was mainly domestic or external. The emphasis was on tracking the main
financial flows, rather than all financial flows. It is in the nature of an exercise such as this that
information seems incomplete, but the objective is to isolate the main flows in order to ascertain the
relevant macroeconomic magnitudes.
An instrument was developed to collect the necessary data from donors and implementing agents,
and results were obtained from a range of financing and implementing agencies. However, it was
not possible to get the data in a desirable consistent form due to variations in the quality of
information across agencies. In particular, many agencies did not keep information in a form which
could enable the disaggregation of spending into the categories required by the project, especially
with regard to domestic versus external spending. Several agencies declined to provide the
requested information, although the eventual coverage was considered extensive enough to provide
useful and representative conclusions.

4

Summary of Results
Information on HIV/AIDS-related spending over the period 2004/05 – 2006/07 was obtained from 19
institutions covering an estimated 60% of total spending over the relevant period. The results
showed that nearly half of the total spending was devoted to treatment, primarily the provision of
ART, at 47% of total spending, with salaries and allowances the next major component at 15%. The
expenditure breakdown in turn suggests that around 60% of total spending was devoted to imported
goods and services, with 40% on domestic goods and services. This significantly reduces the
potential macroeconomic impact of aid inflows.

Demographic Projections
Demographic projections under different scenarios are required for the macroeconomic modelling in
Phase III. In this phase, demographic projections were generated by using the Spectrum model,
which has been specifically developed for modelling the demographic impact of HIV/AIDS, and which
has been used by the Uganda Bureau of Statistics (UBOS) and Uganda AIDS Commission (UAC).
Projections were developed for the period up to 2025 (at least 15 years into the future, as specified
in the ToR), covering the following four scenarios:
• No HIV/AIDS.
• HIV/AIDS without treatment interventions.
• HIV/AIDS with treatment interventions (ART) – How scenario.
• HIV/AIDS with treatment interventions (ART) – High scenario.
The demographic model was calibrated to a 1980 base (i.e., pre-HIV/AIDS). Projections were
produced for a number of variables under the four scenarios:
• Total population (by gender and age group).
• Population deficit due to HIV/AIDS.
• Population growth.
• Number of HIV+ people.
• HIV prevalence.
• Number receiving ART.
• AIDS-related deaths.
• Life expectancy.

Summary of Results
Total Population: by 2002, HIV/AIDS had caused the Ugandan population to be some 6% smaller
than it would have been without HIV/AIDS, while by 2025 the difference would be 9%.
Population Growth: the main impact of HIV/AIDS was felt during the early 1990s, when prevalence
rates were high. As prevalence rates fell owing to High ART use during the late 1990s, the
population is estimated to have risen almost up to the growth rate projected for the “Without AIDS”
scenario.
Impact of ART Provision on the Population: the provision of ART, even in the High ART scenario,
only closes part of the population gap between the “No AIDS” and “with AIDS” scenarios. In the Low
ART scenario, the population in 2025 is only 0.1% higher than in the “No ART” scenario, while in the
High ART scenario, the population is 0.8% higher than in the “No ART” scenario. The reason for the
apparently small impact of ART provision is that a large proportion of the impact on the total
population was felt during the late 1980s and early 1990s, where high HIV-prevalence and death

5

rates had a permanent effect, making the population smaller. The projections also show that unless
ART is widely provided, it pays little demographic dividend.
HIV Population: the number of people infected with HIV is estimated to have peaked at about 1.4
million in 1996, before falling slowly. Without ART, the number of HIV+ people would continue to
decline through to about 2012, following which time it would start to rise again. This reflects a
number of factors. First, even with a constant prevalence rate, if the population is growing then the
number of those infected with HIV will rise. Second, there are indications that the prevalence rate
has been slightly rising, thus reinforcing the upward trend in the numbers of HIV infected people.
With ART, the increase in the numbers of HIV+ people is even more dramatic, especially in the High
ART scenario. The rollout of ART increases the number of HIV+ people, as those who would have
earlier died are now living longer. It is projected that by 2025, there will be 2.2 million HIV+ people
under the High ART scenario, but only 1.8 million in the absence of ART.
Number Receiving ART: The number of people receiving ART continues to rise in both the Low and
High ART scenarios, although much more dramatically in the latter. The number receiving ART in the
High scenario is close to, but somewhat below, the projections contained in the National Strategic
Plan (NSP). This may indicate that the model is under-projecting the number of HIV+ people, or that
the NSP envisages earlier treatment of HIV+ people with ART than the protocols embedded in the
Spectrum model. It is unlikely to reflect a faster rollout of ART in the NSP, as the High scenario in this
model envisages a very rapid rollout of ART.
AIDS-related Deaths: the number of deaths as a result of AIDS is estimated to have been falling since
the late 1990s, which reflects a decline in HIV prevalence. The number of projected deaths owing to
AIDS is highly dependent upon the rollout of ART. Under the High ART scenario, the rapid rollout of
ART initially dramatically cuts the number of deaths due to AIDS, although eventually the number
rises again. In the medium term, ART leads to a significantly reduced death rate and hence improved
life expectancy. Without ART (or in the Low ART scenario), the number of deaths is projected to
decline much more slowly, reflecting only the earlier decline in prevalence.
Life Expectancy: by the late 1990s, life expectancy had fallen to an estimated 44 years, compared to
an estimated 56 years without HIV/AIDS. With time, however, this gap drops, reflecting the decline
in the HIV prevalence rate and, in the High ART scenario, the availability of treatment, which
prolongs the survival times for HIV+ individuals. By 2025, life expectancy is projected to be 60 years
in the “With AIDS” scenarios, compared to an estimated 64 years “Without AIDS.

Analytical (Econometric) Studies on Macroeconomic Relationships
The main channels through which inflows of external donor assistance may impact on the economy
are well known. These include:
• potential exchange rate appreciation due to foreign inflows;
• monetary expansion and inflation if these inflows are taken into reserves (to prevent
exchange rate appreciation);
• fiscal costs and higher interest rates if reserves are sterilised (to prevent monetary
expansion).
While the channels are well understood, there is less information regarding the magnitude of these
relationships. The aim of this study is to quantify these transmission channels.
Two studies were carried out in order to address these issues:

6

• A study of the determinants of Uganda’s exports (principally the real exchange rate, but also
considering other factors as necessary).
• A study of the linkages between aid flows and inflation in Uganda.

Summary of Results
Real Exchange Rate and Exports: This study attempts to quantify the relationship between the real
effective exchange rate (REER) and exports, in order to provide insights into the possibility of an aid
induced Dutch disease effect on the Ugandan economy; which is a current concern given the
increasing aid flows to Uganda towards the control of HIV/AIDS. To accomplish this objective, the
study estimated a model of determinants of six Ugandan exports (coffee, tea, cotton, fish, maize and
flowers) using Vector Error Correction Analysis and quarterly data over the period 1994-2006. The
findings did not indicate a relationship between the REER and total exports. However, the study did
find that the REER would affect specific exports, namely, fish, flowers and cotton, which account for
nearly a quarter of the total exports. Thus, for fish, flowers and cotton, the findings indicate a
possible Dutch disease effect. However, since a possible Dutch disease effect would reduce supply of
some exports, it would have negative implications for poverty reduction in the long run. This
underscores the need to contain appreciation pressures that may arise from aid flows, which have
played a big role in Uganda’s poverty reduction programmes over the last decade. Finally, the
findings indicate that weather (rainfall), terms of trade (TOT), and cost of capital also affect Ugandan
exports, though, to varying degrees depending on the type of export.
Aid Flows, the Exchange Rate and Inflation: This study analyses the impact of aid flows to the
Ugandan economy on prices and REER over the period July 1994 - June 2007, using Vector
Autoregression (VAR). The study found that an increase in aid flow is associated with a long-term
increase in the money supply. However, this does not lead to any long-term increase in prices or to
real exchange rate appreciation, which suggests that the Bank of Uganda’s (BoU) monetary policy
and stabilisation strategy has been successful. In the short run, an increase in aid is associated with
greater volatility in both prices and the REER, which could be damaging to private sector investment.
Moreover, aid dependence leads to high transaction costs (interest costs) through sale of securities
by the monetary authorities. This has negative implications for medium-term fiscal sustainability and
domestic debt sustainability. The policy implication of these findings is, however, not for aid
recipient countries to scale back the level of aid, but rather to strengthen capacity to avoid volatility
in prices, money and financial markets, which is associated with inflows of aid. This could be
achieved through strengthening monetary and exchange rate management by reducing the volatility
of aid flows and increasing their predictability. Finally, given the high short-run adjustment costs
faced by the public sector, a dollar at the margin may have a much higher (social) payoff if it is
transferred directly to the private sector. This could be achieved by retiring domestic debt with the
aim of lowering domestic interest rates and increasing availability of credit to the private sector.

7

Chapter 2: The Impact of HIV/AIDS on
Poverty
1. Introduction
While there has been a considerable amount of research on the macroeconomic impact of HIV/AIDS
in various sub-Saharan Africa (SSA) countries, little analysis of its impact on the household level has
been done. What has been done comprises a mixture of community-level case studies and economy-
wide modelling of the household impact of HIV/AIDS, typically based on actual household survey
data (see Jefferis et al 2007 for a discussion). This chapter aims at establishing the impact of
HIV/AIDS on household poverty in Uganda, by modelling the household income and expenditure
effects. The chapter employs a now well-established methodology first used in BIDPA (2000) and
Haacker and Salinas (2006).
The chapter models the impact of HIV/AIDS on poverty in the absence of widespread public
provision of ART. The intention is to contribute to the debate on the provision of ART, by providing
estimates of one element of the costs of HIV/AIDS (the poverty impact), which can then be
compared with the costs of providing ART.
This chapter first presents information on HIV/AIDS and poverty in Uganda (Section 2), then
introduces the modelling approach (Section 3), and presents results in Section 4.

2. Background: HIV/AIDS and Poverty in Uganda
HIV/AIDS in Uganda
Uganda has been one of the countries hardest hit by the AIDS epidemic over the past 25 years. From
only two known HIV/AIDS cases in 1982, the epidemic grew to reach a cumulative two million HIV
infections by the end of 2000, and it is estimated that of these, around half of them have since died.
It has been estimated that HIV/AIDS has had a direct impact on at least one in every ten households
in the country, including the 884,000 HIV/AIDS orphans (UNAIDS, 2002). The Uganda AIDS
Commission (UAC, 2001) gives similar, but slightly different figures, with a cumulative total of 2.2
million people infected with HIV since its onset, of whom about 800,000 people are estimated to
have died of AIDS; about 1.4 million people were then estimated to be living with HIV/AIDS, of
whom 100,000 were children under 15 years. The UAC gave a much higher figure of over 1.7 million
children orphaned by AIDS. The risk of mother-to-child transmission of HIV (MTCT) was estimated at
15-25% (UAC, 2004)

8

Figure 1: HIV Prevalence Rates, 1988-2005

Source: UAC, MoH
The dramatic reduction in the adult HIV prevalence rate reported in Uganda from around 20% in
1991 to 6.5% in the early 2000s, has been attributed to committed and sustained political
leadership, early intervention, a strong focus on prevention, and a multi-sectoral approach (Okware,
2001). A crucial part in dealing with the consequences is also being played by civil society and
households themselves. The Millennium Development Goal (MDG) for HIV/AIDS aims at halting and
reversing the spread of HIV/AIDS by 2015. While Uganda had achieved this during the 1990s, recent
developments have been less positive, and there are signs that the HIV-prevalence rate has been
rising. For instance, the target in 2002 was to reduce HIV/AIDS prevalence to 5% by 2005, but it
remained as high as 6.3% among the adult population (15-59 years) according to the 2004/2005
Sero-Behavioural Survey.
Despite doubts over the accuracy of some of the historical estimates of the prevalence rates as a
result of constraints on HIV/AIDS reporting and uneven coverage of sentinel sites, there is no doubt
that Uganda has succeeded in achieving a significant reduction in HIV prevalence over the last
decade. This partly reflects the deaths of many people infected by HIV, but also a marked reduction
in new transmissions.
Regional differences in HIV prevalence can easily be noted (see Figure 2). There are especially high
prevalence rates in Central and Kampala districts (which could be due to relative economic
prosperity) and the North Central region. The latter could be attributed to refugee settlements and
internally-displaced persons (IDPs) in the area; as noted in the Poverty Eradication Action Plan
(PEAP) (MoFPED, 2004). Key challenges include a relatively high HIV/AIDS prevalence in refugee-
affected regions. Displacement and migrations from other countries increases the host communities’
exposure to HIV/AIDS and other STDs. Redundancy, trauma, poverty and ignorance also contribute
to the spread of such diseases.

9

Figure 2: Adult HIV Prevalence Rates by Region

Source: MoH, 2006
Uganda’s human resource planning has not been systematic and most sectors do not have a
systematic way of accounting for the effects of HIV/AIDS on their staffing or of predicting future
requirements of staff as a result of HIV. Also noted is that the effect of HIV/AIDS on productivity in
the public service has not been properly estimated.

Provision of ART
As a matter of policy, the GoU has decided to provide free treatment for HIV/AIDS through the
provision of ART. This is to be accompanied by close monitoring of adherence to avoid the
emergence of drug resistance. Recent data suggest that there has been an increase in the number of
people receiving ART from 17,000 in 2003 to about 110,000 by mid 2007 as shown in Figure 3 below
(UAC, 2007; MoH, 2007).

10

Figure 3: Total Number of People on ART

Source: MoH, 2007
Besides UAC planning to reduce the incidence rate of HIV by 40% by 2012, the national target is to
increase equitable access to ART by those in need to 240,000 by 2012 (UAC, 2007).
The objective of scaling up ART roll-out further requires large-scale capacity-building activities,
including additional training of health care workers and development of a stable health care
infrastructure. This may further demand the re-allocation of public funds from other sectors.
The provision of ART falls under the government’s general health financing programme, which has
the following as the general guiding policy: “To develop and implement a sustainable, broad-based
national health financing strategy that is geared towards:
i) Ensuring effectiveness, efficiency, and equity in the allocation and utilisation of
resources in the health sector consistent with the objectives of the National Poverty
Eradication Action Plan.
ii) Attaining significant additional resources for the health sector and focusing their use on
cost-effective priority health interventions.
iii) Ensuring full accountability and transparency in the use of these resources.
The government plans to archive this through:
a) Progressive increases in the financing of the sector;
b) Focusing the use of public resources on the most cost-effective health services while
protecting the poor and most vulnerable population and considering all gender-related
health care concerns;
c) Developing and supporting alternative financing schemes such as user fees and health
insurance without discriminating against the poor and vulnerable groups; and
d) Promoting the growth of private sector health initiatives.
At a household level, HIV/AIDS infection has major implications for household expenditures and
vulnerability to poverty. The cost of AIDS treatment (when paid out-of-pocket) competes with other
crucial expenditures, such as food, shelter and, educational expenses. Even if treatment is
subsidised, there are other costs associated with treatment such as cost of transport to the

11

distributing centre for treatment, costs of other medications, potential loss of income during times
of illness, and diversion of funds toward healthcare (PEAP 2004).

Poverty Levels
The estimated national incidence of poverty fell dramatically from 56% in 1992 to 35% in 2000,
before rising again to 38% 2002. The most recent estimate for 2005/2006 is a national headcount
poverty rate of 31%. By 2017, the target is to reduce the poverty rate to 10%. A number of policy
interventions, including the Poverty Action Fund (PAF), have significantly improved social service
delivery and its impact. However, less investment has gone to the productive sectors, particularly
agriculture, where the majority of the poor derive their livelihood. Rural communities remain poorer
than the urban population, with significantly higher poverty rates than in urban areas. Inequalities
between socioeconomic groups and regions also persist. In 2002/2003, the overall poverty rate was
estimated at 38%, with the lowest rate in the Central region and the highest in the Northern and
Eastern regions (PEAP 2004).
Figure 4: Trends in Poverty and Inequality

Source: PEAP 2004, Uganda Human Development Report 2007 and UNHS 2005/2006
Despite the reduction in poverty levels, there has been a marked increase in inequality since 1997.
The Gini coefficient, which measures inequality, rose from 0.36 in 1992/93, to 0.43 in 2002/03
although it has fallen to about 0.41 in 2005/06. The reasons for this pattern are thought to include a
slowdown in agricultural growth during the past years, declines in farmers’ prices, insecurity, high
population growth rate and morbidity related to HIV/AIDS.
HIV/AIDS remains the leading cause of death within the most productive age ranges of 15-49. With
HIV/AIDS prevalence rates stagnating between 6% and 7%, even though down from 18% in the mid-
1990s, a further reduction in morbidity is dependent upon large-scale rollout of ART. Co-ordinated
multi-sectoral action is required to reverse these trends, and mitigate the impact of HIV/AIDS (PEAP
2004).
In trying to identify whom the poor are, PEAP 2004 concerned itself with four particular issues:
regional inequalities; gender; occupational structure; and other disadvantaged groups.
While most parts of the country shared in the benefits of growth between 1992 and 2000, the North
was left behind. The proportion of people in the North below the poverty line fell from 72% in 1992

12

to 60% in 1997/98, but rose again to 66% in 2000, and in 2005/06 Uganda National Household
Survey (UNHS), it was estimated to be 60.7%.
The second-poorest region, the East, also suffered a significant deterioration, partly because of
distress migration.
Figure 5: Regional Poverty Rates

Source: PEAP 2004, UNHS 2005/2006

HIV/AIDS and Poverty
Impact of HIV/AIDS
In Uganda, HIV/AIDS policies are emphasising the mitigation of the impact of HIV/AIDS and the
universal provision of ART. It has been recognised (in PEAP 2004) that AIDS is a development issue:
• there is an interaction between HIV and poverty;
• at national level, HIV/AIDS robs sectors of both skilled and unskilled labour;
• it diverts scarce resources that could have been used productively in other sectors;
• AIDS increases absenteeism from work due to frequent illness of staff or/and nursing of sick
family members leading to decreased productivity;
• the impact of HIV on labour supply has affected agricultural growth in some regions;
• there is a sharp increase in the proportion of investors reporting that AIDS is a constraint.
The epidemic also affects public sector service delivery, household savings and the intergenerational
transmission of knowledge, and imposes a greater burden on the elderly while reducing their
economic security. By killing primarily young adults, AIDS does more than destroy the human capital;
it also deprives their children of the requirements (parents’ care, knowledge, and capacity to finance
education) to become economically productive adults.
At the micro level, HIV/AIDS affects particular social groups like orphans and vulnerable children
(OVC), women, refugees and IDPs, who have been especially hit by the epidemic due to their
disadvantaged position and low incomes. This is increasing the risk of children becoming street
children, or a target for abuse and exploitation.

13

Household Level: HIV/AIDS, Poverty and Inequality
The existing literature suggests that one of the greatest impacts of the HIV/AIDS epidemic is felt at
the household level, where socio-economic factors combine with socio-cultural and epidemiological
variables to influence prevalence (SSRC, 2004). It is the household unit that carries the greatest
burden. Since socio-economic indicators, such as poverty and inequality, are both consequences and
determinants of HIV/AIDS, they can interact with the epidemic at a household level to perpetuate a
vicious downward cycle towards greater indigence.
Several studies have suggested that poverty increases susceptibility to contracting HIV/AIDS through
several channels, including increased migration to urban areas; limited access to health care,
nutrition and other basic services; limited access to education and information; sexual exploitation
and gender inequality (Bloom et al, 2004).
Research evidence has also shown that the epidemic’s influence on household living conditions
derives in great part from the virus’ specific demographic effects. Since HIV/AIDS is distinct from
other diseases because it strikes prime-aged adults (the most productive segment of the economy
15-59 year old population), it changes the structure of the population (Barnett and Whiteside, 2002).
HIV/AIDS makes the breadwinners fall ill and die, destroying the much-needed skills and depriving
children of their parents. Barnett and Clement (2005) point out that the key to the social and
economic impact of HIV/AIDS is that it is a slow moving virus and as a result it can affect three
human generations.
The principal economic impacts experienced by affected households are loss of available income, as
working adults fall ill or die or have to stop work to look after children and/or the ill as well as
additional expenditure on health care and funerals (UNAIDS, 2004). Other effects include depletion
of household assets (due to increased health expenditure, consumption needs and labour losses),
lower productivity of subsistence labour and reduced availability of food. School enrolment may also
decrease, as children are forced to dedicate time to labour and care-giving.
In a survey of 771 AIDS-affected households throughout South Africa, Steinberg et al (2002)
documents the impoverishment and burden of care for family members. The epidemic deepens
poverty among the already poor through loss of income and medical care costs, which absorb up to
one third of household income. Children’s schooling is also disrupted, especially among girls. This
study also reveals a growing strain on extended family networks as households often send their
children to live elsewhere, most often with relatives, worse still the already ageing grandparents.
Similar dynamics are described in Bachmann and Booysen’s (2002/04) 18-month longitudinal study
of rural and urban households in South Africa’s Free State Province. The baseline study (2001-2002)
finds that affected households are poorer than non-affected households, regardless of the poverty
measure used. The incidence, depth and severity of poverty were worse among affected
households, particularly among those who experienced illness or death. Some new findings of the
follow-up studies are the insignificant differences in the impact on rural and urban households and
the decline in income of unaffected households. The latter phenomenon suggests that the effects of
the epidemic are not limited to “infected” households, but are giving rise to deepening poverty in
the wider community.
Another survey carried out in South Africa (Oni et al, 2002), in the Limpopo Province, provides
further evidence of how HIV/AIDS worsens poverty among households already living below the
poverty line. One empirical result is that income received by affected households during the year
2000 was approximately 35% lower than that received by unaffected households while per capita

14

monthly income for the average affected household was about 31% lower than that of unaffected
households. The study brings to light changes in household expenditure patterns where health and
medical care as well as transportation and funeral expenditure increased among affected
households, while spending on education, housing and remittances was reduced. For example,
affected households increased their transportation costs by 4.7% and reduced expenditure on
education by 7.3% and housing by 11.5%.
In a paper published in 2004, Wyss et al attempt to ascribe a value to the household level economic
costs of HIV/AIDS described above. Their fieldwork in Chad, one of the poorest countries in the
world, confirms that for most households, especially in the low-income settings, the consequences
of AIDS are disastrous. Costs attributable to the epidemic during the period of illness up to death
represent more than four times the annual Gross National Product (GNP) per head in Chad.
Productivity losses make up 28% of total costs, while 56% of costs are on health related expenditure
and 16% on funeral expenses.

Long-term Household Costs of HIV/AIDS
There are also more indirect and long-term repercussions of the epidemic on households, that are
not immediately apparent. Some go beyond the economic sphere, such as grief and increasing
stress, which can negatively influence the psycho-social state of children. But there are also potential
long-term economic costs. One intergenerational effect is that of diverting household resources
from long-term assets to meet short-term needs, which influences household savings and
investment decisions (Greener, 2004; ILO, 2004). Another loss is that of human capital, as fewer
household resources – time, money, care, etc – are directed to children’s mental and emotional
development.

Household Coping Strategies
When households that already live on the margins of survival are forced to absorb the ‘shock’ of
HIV/AIDS, there is little else they can do but struggle to go on with whatever means possible. Coping
strategies of affected households include utilising household savings, risk pooling and selling
household goods. In some cases, families have no alternative but to sell productive assets (e.g. land,
buildings, livestock, business stock, tools, etc), getting children out of school, and relocating from
urban to rural areas more especially when the bread winner is the one that has died, thus further
frustrating income generation potential and the possibility of recovering some of the losses incurred
(Jayne, 2003/04). In this way, the situation of poverty is intensified and there is little opportunity for
upward socio-economic mobility. Families rarely recuperate their initial level of economic well-
being.
Robalino et al (2002) report that in Middle Eastern and North African countries, informal solutions to
manage risks are diverse, ranging from family support and kinship ties to religious charitable
organisations, but research has shown that they are usually insufficient to hedge against adverse
shocks. Wyss et al (2004) find that AIDS cases in Chad rely more often on borrowing and selling of
household assets for treatment, compared to non affected households. Across all households,
income and savings are the most important sources for covering treatment costs.
Nampanya-Serpell’s research among urban and rural households in Zambia (2002) finds that in the
urban sample, the worst affected families are those in which the major breadwinner was the first
parent to die. These families experienced a sharp drop in income and in most cases were forced to
move out of their original home. Those who own a home and rent out part of it, those with an

15

educated adult female employed in the formal sector and those with wealthier relatives who can
take in orphans are most protected from hardship.

3. Modelling the Household Level Impact of HIV/AIDS
Methodology and Previous Studies
The most comprehensive effort to model the impact of HIV/AIDS on household poverty was carried
out by Salinas and Haacker (2006). This modelled the impact in four SSA countries: Ghana, Kenya,
Swaziland and Zambia, using methodology first used in Botswana by BIDPA (2000) and Greener,
Jefferis & Siphambe (2000), and more recently by Jefferis et al (2006). The basic approach is to use
pre-existing household survey data on income and expenditure, and then to hypothetically model
the impact of HIV/AIDS on each household through income and expenditure effects, and hence on
households’ poverty status. HIV/AIDS is allocated randomly across individuals within households,
with the probability of each individual becoming HIV+ matching prevalence data according to
demographic, economic and social characteristics. This is most easily done if household-based Sero-
Prevalence Survey data is available, but if not sentinel survey data can be used1.
Once HIV status has been randomly assigned to individuals in the sample, based on the respective
individual’s socio-economic characteristics and the information available on HIV prevalence, this is
then aggregated to household level. It is then used to simulate the impact of HIV/AIDS on income
and consumption per capita, income distribution, and poverty rates.
A number of assumptions are required for this simulation and analysis. For HIV+ individuals, it was
assumed that they live an average of ten years with the infection. It was further assumed that
HIV/AIDS will impact economically on households by both increasing its required expenditures and
by reducing its income through morbidity and mortality. The additional HIV-related expenditures
assumed were health-care (assumed to be 25% for urban and 50% for rural households of incomes),
and funeral expenditures (equivalent to four months of household expenditure). These expenditures
were taken as an addition to their minimum expenditure defined in the poverty line.
This study also assumed that when a household member dies, his or her income is lost, and the
average household income declines correspondingly. If a household member without income dies,
the remaining income is divided among fewer household members. It is also assumed that there is a
reduction of 15% in the income of any worker in the household who is HIV-infected and in the last
two years before HIV-related death.
The HIV prevalence rates in the four countries show considerable variation, from 2.1% in Ghana to
31.4% in Swaziland. Salinas & Haacker’s base case scenario results show that HIV/AIDS has a
considerable negative impact on poverty and inequality, although it depends on exactly which
measure is used. The greatest negative impact is on poverty measured at the US$ 1/day poverty line,
whereas the impact on the US$ 2/day poverty line is much less, presumably reflecting the
distributional characteristics of the disease. Also, in the countries with relatively low prevalence
rates (below 10%, in Ghana and Kenya), average per capita incomes fall by less than 1%, whereas in
the high prevalence countries the reduction in per capita incomes is much higher.

1
Sentinel survey data were used for the first Botswana study and for Swaziland, while Sero-Prevalence
Survey data were used for Ghana, Kenya, Zambia and the second Botswana survey.

16

Table 1: Impact of HIV/AIDS on Poverty and Inequality in Selected Countries
Ghana Kenya Swaziland Zambia
Year 0 Year 10 Change Year 0 Year Change Year 0 Year 10 Change Year 0 Year 10 Change
(%) 10 (%) (%) (%)
Adult HIV 2.1 6.7 31.4 15.6
prevalence (%)
Income per 2.49 2.47 -0.55 2.80 2.78 -0.90 2.87 2.65 -7.51 1.75 1.57 -9.95
capita (US$)
Poverty 66.38 66.38 0.003 58.92 60.53 2.7 59.90 64.00 6.9 79.17 81.48 2.9
headcount (US$
2/day,%)
Poverty 34.15 35.74 4.6 23.46 25.80 10.0 23.28 33.00 41.8 51.04 57.52 12.7
headcount (US$
1/day,%)
Gini coefficient 53.85 54.49 1.2 48.43 50.41 4.1 50.63 55.74 10.1 47.77 53.23 11.4
Source: Salinas & Haacker, 2006

The earlier study of the impact of HIV/AIDS on poverty and inequality in Botswana (Greener et al,
2000) found that HIV/AIDS would increase the poverty rate by 6 percentage points (pp) (from 37.7%
to 43.7%). This 16% increase in poverty is consistent with the numbers for the high prevalence
countries in the Salinas & Haacker study. The more recent Botswana study (Jefferis et al, 2006)
found a smaller increase in poverty of around 3pp (from 33% to 36%). This smaller increase of 9% in
the poverty rate partly reflected a lower HIV prevalence rate at the time of the later study, which
itself stemmed from both success in containing HIV/AIDS and improved data (using household Sero-
Prevalence Survey data rather than sentinel survey data)2.

Approach followed in this Study
In an effort to estimate the magnitude of the impact of HIV/AIDS on the household level, the key
focus was its impact on poverty. The following procedure was taken to arrive at these estimates:

Simulating HIV/AIDS
The analysis made use of person-level and household data from the 2005/06 UNHS. Using the
HIV/AIDS Sero-Behavioural Survey data of 2004/05, each person in the UNHS data was assigned an
HIV status in accordance to the age, sex, and region; level of education attained; as well as
employment and marital status of the individual. This resulted in a pattern of infection which
resembled very closely that observed in the Sero-Behavioural Survey. The person-level information
was then aggregated back to household level in order to simulate the household impacts. Using
certain assumptions about costs of HIV/AIDS to the affected households, the income and
expenditure effects, we simulate the impact of HIV/AIDS on poverty, income and expenditure per
adult equivalent under different scenarios. Note that there can be more than one HIV+ person in
each household.

Key Assumptions
This analysis combined the most recent sources of information about household structure, sources
of income and expenditure patterns as well as HIV prevalence. The validity of the analysis rests on a
number of key assumptions, as described below.
In the absence of ART, a person will die within 10 years from the time he/she gets infected with HIV.
Ten years was taken as an average, since some people with low incomes, poor dietary habits and
general poor health conditions may die before the 10 years, while others with better access to

2
The 2004 Botswana AIDS Impact Survey measured a 24% adult HIV prevalence rate, compared to figures of
around 35% from earlier sentinel surveys.

17

health facilities and food may live with HIV for more than 10 years. There are also differences in the
way individuals respond physiologically to infection.
With or without HIV, household composition, structures and income sources change. Natural factors
of population growth (death and birth) will still bring about population increase, and new
households would be formed as people marry and get married, people will be growing and those
currently not working will have started to work because of their advancement in age and this will
add to household income while others will die leading to loss in income. All these have been
assumed to roughly cancel out the impact of HIV/AIDS on household income distribution.
Within ten years, changes in the population distribution by age and sex are assumed to be
insignificant, since changes in overall demographic structure are comparatively slow. The analysis
here essentially assumes that the population structure will be in a steady state, apart from the
effects of HIV/AIDS. This assumption isolates the impact of HIV/AIDS from the impact of other
demographic changes taking place at the same time.
The economic impact of HIV/AIDS on the household is assumed to arise out of increased household
expenditure and reduced income due to morbidity and mortality.
With regard to household expenditure, it is assumed that a person will begin to develop AIDS
symptoms in his/her eighth year from the time of infection with HIV3. Hence, more frequent illnesses
may be experienced, and this will increase medical and related household expenses (these include
direct medical expenses such as consultations, laboratory tests, medication, hospital admission as
well as indirect expenses such as transport costs, special dietary requirements, etc). However, the
total expenditure in actual terms may depend on the household’s income, although the proportion
may fall in a given acceptable range. In Uganda where the health insurance is at its very minimal,
most of these costs are financed from household’s own sources, and some households may have to
sell off assets and properties like land in order to meet these costs. These costs are assumed to
increase by 50% in rural areas and 25% in urban areas due to the easiness or difficulties faced in
accessing health services and differences in actual incomes4.
The second associated cost is funeral-related expenses when a person dies of AIDS. These costs,
which include feeding of many mourners for several days, and purchasing of other requirements like
caskets, making announcements, transport, etc, tend to be high. Also, the actual amount spent will
depend on the social status and level of income of the household that has lost a member, although
the proportion may fall in a given range. It was assumed that funeral expenses would be equivalent
to four months of household expenditure5.
Since these expenditures are almost indispensable to the household, we reflect them as additions to
their minimum expenditure or Poverty Datum Line (PDL) (following Salinas and Haacker, 2006). The
approach taken here was to add these additional expenditure requirements to the package of basic
needs of a household. In other words, there is an addition to the PDL of households affected by
HIV/AIDS. The effect of this is to re-define the level of income which constitutes poverty and thus a
family affected by HIV/AIDS is more likely to be classified as poor, due to these additional expenses.

3
Again, this does not mean that all HIV+ individuals will develop AIDS in the eighth year after infection.
Some will develop it earlier, others later. The figure is intended as a representative average, and has been
used in other similar studies.
4
These parameters are the same as those used in Salinas and Haacker (2006), and derived from the detailed
household study of the impact of HIV/AIDS in South Africa (Steinberg et al, 2002).
5
As in Salinas & Haacker (2006).

18

These are short-term expenses, which apply in the period up to the tenth year when the HIV+
individual is assumed to die. The longer-term impact is therefore different, as the household no
longer has to meet their expenditure requirements. This has the opposite effect to the additional
expenditures described above, and makes it less likely that an HIV affected household will be
classified as poor (as the household is now smaller and has lower expenditure requirements, and
household income has to be spread across fewer members).
However, expenditure effects are not the only effects on household poverty levels. There are also
income effects, which are discussed below.

Income Effects
The incomes of individuals and households in general are also affected by HIV/AIDS through higher
mortality and morbidity. In the short-term, while one or more household members are sick,
household income will be reduced, because a breadwinner may be unable to work due to illness or
the need to care for other household members67.
In the long-term, the worst scenario that the household can face is that a breadwinner or an income
earner dies of AIDS. This means that the income of the household goes down, and even though it is
divided between fewer members, the per capita income of the household is likely to be lower
because of the lost income. Some family members may respond by looking for jobs, and if they are
successful this may lead to some or all of the lost income being replaced. Salinas and Haacker (2006),
contend that having an efficient labour market which enables the unemployed to find jobs relatively
quickly when others die of AIDS plays an important role in offsetting the negative impact of poverty
on HIV/AIDS. However, in Uganda, this is likely to be a less important channel. The vast majority of
the workforce is already occupied and unemployment is low, at 1.9% of the labour force in 2005/06,
hence there are few people who can readily move into vacancies created when others die8. This is
especially so in the case of skilled or professional jobs, when it is difficult to have the same skills, and
even then, a person with skills does not wait for someone to die in order to look for a job. The
majority of Ugandans (74% of the labour force) are employed in the agricultural sector, and reduced
labour availability is likely to lead to reduced agricultural output and incomes, although there may be
some scope for those who are underemployed (12.6% in the rural areas in 2005/06) to make up for
some of the loss.
It is therefore assumed that a person who dies is hard to replace, and the lost income is foregone for
that household. Other incomes earned thereafter would meet the day to day household
expenditures since those households still have to live within limited income.
By comparison, if a household member without income dies, the income is simply divided amongst
fewer household members, hence such a household is less likely to be poor and the per capita
income of the household may increase.

Indicators
The analysis calculates the values of a number of key indicators of poverty and inequality before and
after the 10 year period. The indicators considered were as follows:

6
This study models only the income loss from breadwinners who become sick, due to lack of information
regarding care relationships within the household.
7
The study by Steinberg et al in South Africa found that two-thirds of AIDS-affected households experienced
a fall in income.
8
Data from UBOS Report on Labour Market Conditions in Uganda, December 2007.

19

Poverty levels are measured in the same way as they were in the UNHS 2005/06. Households are
classified as poor if their consumption per adult-equivalent is less than the relevant adult-equivalent
poverty line. We use household poverty, which is the percentage of households below the poverty
line9.
The household per-capita income is the household disposable income reported in the UNHS
2005/06, divided by the number of household members. It is important to note that this is not the
same as per-capita Gross Domestic Product (GDP), and should be expected to be substantially lower.
It was also adjusted for the HIV/AIDS cases using adjusted household income.
Another useful statistic is the income dependency ratio. This is the average number of people
(within a household) who are supported by each household member who is employed and earning
an income. This is usually considered to be a sensitive indicator of household poverty, and of the
vulnerability of a household to the loss of an income earner. It was adjusted also for those with AIDS
cases in the household given that their ability to earn an income is greatly hindered by frequent
illnesses.

Poverty Data
The UNHS dataset includes information on both households and the individuals within the
household. In order to relate the individual and household data with regard to consumption, income
and poverty levels, it is necessary to designate all household members with an adult-equivalent
status. The adult-equivalent conversion factors specify what proportion of an adult’s consumption
level is required by household members of different ages and genders. The conversion factors used
to derive household consumption levels and poverty lines are specified below:
Table 2: Adult Equivalent Conversion Factors

Age group Male Female
<1 yr 0.27 0.27
1 yr 0.39 0.39
2 yrs 0.45 0.45
3 yrs 0.52 0.51
4 yrs 0.57 0.56
5 yrs 0.62 0.60
6 yrs 0.67 0.63
7 yrs 0.71 0.67
8 yrs 0.75 0.70
9 yrs 0.79 0.74
10 -12 yrs 0.87 0.78
13-15 yrs 0.97 0.83
16-19 yrs 1.02 0.77
20+ yrs 1.00 0.73
Source: WHO (computed from the data of the UN Food and Agriculture Organisation. Energy and protein
requirements report of joint FAO, WHO expert group)

The following PDL (Table 3) estimates were used in this modelling exercise. These figures were
compared with the adjusted household consumption data (UNHS 2005/2006) to derive the poverty
status for each household.

9
Note that poverty status (below the poverty line or not) is measured using consumption expenditure,
which is generally considered to me more reliable than income. When the impact of income changes (due
to HIV/AIDS) is modeled, it is assumed that consumption changes by the same amount as income.

20

Table 3: Adult-equivalent PDL Estimates as used in 2005/2006 UNHS

Region Shs
Central rural 21,322
Central urban 23,150
Eastern rural 20,652
Eastern urban 22,125
Western rural 20,872
Western urban 21,800
Northern rural 20,308
Northern urban 21,626
National (average) 21,135
Source: UBOS

The poverty rates derived in this way are shown in Table 4 below. The overall poverty level at
household level was about 27%. Northern Uganda had over half of its households (52%) classified as
poor, followed by Eastern Uganda (29%). The proportion was lowest in Central region (14%) and
Western region (16%)10.
Table 4: Poverty Rates by Region

Region Poverty rate
Central 0.14
Eastern 0.29
Northern 0.52
Western 0.16
Overall 0.27
Source: Based on the UNHS 2005/2006 data

4. Results
In estimating the impact of HIV/AIDS on household poverty levels, we considered a range of
different impacts:
Short-term impact:
• Health and related costs.
• Funeral costs.
• Income effect.
Long-term impact:
• Income effect.
• Household composition effect.

Short-term Impact
Impact of Health and Related Costs
Under this scenario, rural and urban households were assumed to be facing different levels of
additional cost burden. Following the precedent of similar exercises elsewhere, it was assumed that

10
These figures are slightly lower than those reported in UNHS 2005/06 (UBOS 2007). This exercise was
unable to replicate exactly the UBOS poverty calculations. However, the main purpose of this exercise is to
model the change in poverty rates, not the absolute levels.

21

rural households will spend an equivalent of 50% of their monthly household consumption
expenditure on HIV/AIDS-related health and other needs while the urban households would spend
an equivalent of 25%. These costs were only applying to individuals that had lived with HIV for eight
to 10 years, as this is when the impact of the illness becomes most severe. As discussed above, the
impact was incorporated through raising the appropriate poverty line for the household.
Taking account of this impact only, HIV/AIDS-related health and other costs made the household
poverty level to increase from 26.8% to 27.8%, i.e. an increase of 1pp in the poverty level. This
effect was generally felt more in rural areas (1.2pp) than in urban areas (0.3pp). Also, the
households in Northern region behaved differently from the rest of the regions. This could be
attributed to the political disturbance that influenced their rural/urban settlement patterns other
than the social-economic factors that operate in other regions. The Western rural households
experienced a 1.7pp increase in poverty rates due to HIV/AIDS related health costs.
Table 5: Changes in Poverty Level: Health Costs Impact

Region % Without HIV/AIDS % With HIV/AIDS Change
Rural Urban Overall Rural Urban Overall Rural Urban Overall
Central 17.7 4.3 14.3 19.1 4.7 15.4 1.4 0.4 1.1
Eastern 33.4 11.8 28.8 34.6 12.1 29.7 1.1 0.2 0.9
Northern 57.5 33.4 52.4 58.2 33.4 53.0 0.7 0.0 0.6
Western 18.5 7.8 15.9 20.3 8.3 17.4 1.7 0.5 1.4
Overall 31.0 12.8 26.8 32.2 13.1 27.8 1.2 0.3 1.0

Impact of Funeral Costs
The funeral costs scenario assumed that people with HIV/AIDS will not die until the 10th year of living
with the virus. Funeral costs were estimated to be an equivalent of four months household
consumption expenditure. Having this effect and ignoring other factors gives a smaller impact on
household poverty. This could be attributed to the fact that funeral expenses are a one-off expense
which are relatively small compared to overall household consumption. Generally, the increase in
number of households moving into poverty as a result of funeral expenses only was less than 1pp for
both rural and urban households. The social ties and community responsibility to support
households that have lost family members protects such households from selling family assets such
as land in order to cater for burial expenses. The shouldering of the burial costs tends to rest on
friends and relatives of the bereaved household.
Table 6: Changes in Poverty Level: Funeral Costs Impact

Central 17.7 4.3 14.3 17.7 4.5 14.3 0.0 0.2 0.0
Eastern 33.4 11.8 28.8 33.6 11.8 28.9 0.1 0.0 0.1
Northern 57.5 33.4 52.4 57.6 33.4 52.5 0.1 0.0 0.1
Western 18.5 7.8 15.9 18.6 8.0 16.1 0.1 0.2 0.1
Overall 31.0 12.8 26.8 31.0 12.9 26.9 0.1 0.1 0.1

Income Adjustment Effect
In this scenario, the assumption was that in the early years of HIV, the person is still effective in
his/her work and there is no negative effect on income. The efficiency, however, begins to decrease
in the eighth year to about 85%, then to 60% in the nineth year and finally to 10% in the 10th year of

22

HIV. This level of inefficiency is translated into reduced income by the same proportion. The lost
income simply translates into reduced available money to consume by the same household.
Assuming other factors remain constant, then the households with an HIV+ working member who
has developed AIDS will face this challenge that will push it into poverty or more poverty if it was
already poor.
The income adjustments were able to increase poverty levels among households in Central and
Western regions by about 1pp, while those in North and Eastern region did not face such an impact.
This could be attributed to the fact that the probability of a household member being in a gainful
employment was higher in Central and Western region than in the Northern and Eastern regions.
This meant that having an AIDS person that is working was more likely to be in Central and Western
region. This further illustrates the regional imbalances in income and poverty distribution.
Table 7: Changes in Poverty Level: Income Adjustment Effect

Central 17.7 4.3 14.3 18.4 5.0 15.0 0.7 0.7 0.7
Eastern 33.4 11.8 28.8 33.7 12.1 29.1 0.3 0.2 0.3
Northern 57.5 33.4 52.4 58.0 33.7 52.9 0.5 0.3 0.4
Western 18.5 7.8 15.9 19.1 8.3 16.5 0.6 0.5 0.6
Overall 31.0 12.8 26.8 31.5 13.3 27.3 0.5 0.5 0.5

Income and Expenditure Effects Combined
In this scenario, the health and funeral costs effects were incorporated through a higher poverty line
while the income adjustments affected disposable income for household consumption.
Incorporating all these factors in the model made the household poverty level increase from 26.8%
to 28.2%, which is a 1.4pp increase. This was more felt in Central and Western regions, with much
smaller effects in the Eastern and Northern regions, as can be noted in Table 8. Rural households
were more effected (1.6pp) than urban households (0.9pp). It is clear that the simple headcount of
households that are poor or not shows that regions with higher poverty rates (Northern and Eastern
regions) experienced less increase in head count of poor households than those with lower poverty
rates (Central and Western regions). This will be clarified further in the estimations of the poverty
gap (P1) and the severity of poverty (P2) which help to measure the depth of poverty. The overall
magnitude of the changes in poverty levels (5.2% nationally) is comparable with that found by
Salinas and Haacker (2006) for Kenya, with a similar HIV prevalence level, which varied between
2.7% (for a US$ 2/day poverty line) and 10% (for US$ 1/day).
Table 8: Poverty Levels Incorporating Income Adjustments, Health and Funeral Costs Effect

Region % Without HIV/AIDS % With HIV/AIDS
Rural Urban Overall Rural Urban Overall
Central 17.7 4.3 14.3 19.9 6.0 16.4
Eastern 33.4 11.8 28.8 34.7 12.1 29.8
Northern 57.5 33.4 52.4 58.3 34.0 53.2
Western 18.5 7.8 15.9 20.5 8.7 17.7
Overall 31.0 12.8 26.8 32.6 13.7 28.2

23

Table 9: Changes in Poverty Levels (percentage points and %)

Region Change (pp) Change (%)
Rural Urban Overall Rural Urban Overall
Central 2.3 1.7 2.1 12.4% 39.5% 14.7%
Eastern 1.3 0.2 1.0 3.9% 2.5% 3.5%
Northern 0.8 0.6 0.7 1.4% 1.8% 1.5%
Western 1.9 0.9 1.7 10.8% 11.5% 11.3%
Overall 1.6 0.9 1.4 5.2% 7.0% 5.2%

Long-term HIV/AIDS Impact
As already pointed out in the assumptions, the major factor considered was that all people who are
HIV+ will die by their 10th year, and households will therefore lose incomes earned by such members
who are working. In the long-run, the impact of funeral and health costs, which are transient, will
have passed. However, the effect of lost income from working household members who die will be
permanent. Also considered was the fact that the size of households will be reduced by the number
of deaths in that household, and hence the total adult equivalent value of that household will
change.
Taking account of these combined effects (income and household composition), poverty rates are
estimated to rise from 26.8% to 27.3%, an increase of 0.5pp, with similar changes in urban and rural
households.
Table 10: Long-term Changes in Poverty Level: Income Adjustments due to Death of HIV+Iincome
Earners

Region % Without HIV/AIDS % With HIV/AIDS % Change
Central 17.7 4.3 14.3 18.4 5.0 15.0 0.7 0.7 0.7
Eastern 33.4 11.8 28.8 33.7 12.1 29.1 0.3 0.2 0.3
Northern 57.5 33.4 52.4 58.0 33.7 52.9 0.5 0.3 0.4
Western 18.5 7.8 15.9 19.1 8.3 16.5 0.6 0.5 0.6
Overall 31.0 12.8 26.8 31.5 13.3 27.3 0.5 0.5 0.5

HIV/AIDS Impact on Poverty Gap and Severity Indices
The analysis of the impact of HIV/AIDS on household poverty did not stop at the headcount indices
showing the number who are poor. It proceeded to assess the poverty gap and the severity of
poverty created by the pandemic. This is because the headcount fails to show the depth and severity
of poverty created under the above scenarios. It shows the number of households that were falling
below the poverty line under the different scenarios, but does not tell us how poor the poor are, as
it remains unchanged when the poor become poorer under the different scenarios. Ideally, poverty
measures should fall as the living standards of poor households rise or when income is transferred
from a non-poor household to a poor household (even if the poverty status of both remain
unchanged). The modelling therefore looked at the distribution of income among the poor as well as
the incidence of absolute poverty.
The additional poverty indicators calculated were P1 and P2, for the different scenarios discussed
above. P1 is the average poverty gap in the population (the amount by which household
consumption falls below the household poverty line) expressed as a proportion of poverty line (PDL).
For example if P1 = 0.3, it means that the average aggregate deficit of the poor relative to the

24

poverty line, when averaged over all households (poor and non poor households), represents 30% of
the poverty line.
P2 on the other hand is a distributionally sensitive index that can detect income (or expenditure as a
proxy of income, as in our case where expenditure was used) among the poor. P2 measures the
square of the poverty gap, and hence gives more weight to households far below the poverty line.
This measure would increase in the event of a transfer of income from a poor to a less poor
household, even if both are below the poverty line, whereas P1 would be unchanged by such a
transfer.
Looking at the different short-term impact scenarios, there is clear evidence that the poverty gap
has increased due to HIV/AIDS and severity of poverty has also worsened. However, the impact has
been different for different regions and whether the household is in a rural or urban area.
Table 11: The Effect of Different Scenarios on the Poverty Gap (P1) and Severity (P2)

No HIV All factors Health costs Funeral costs Income
adjustment
P1 P2 P1 P2 P1 P2 P1 P2 P1 P2
Central rural 5.27 2.19 5.34 2.25 5.31 2.22 5.28 2.20 5.30 2.21
Central Urban 1.20 0.57 1.27 0.65 1.22 0.59 1.22 0.59 1.25 0.62
East rural 9.89 4.13 10.13 4.35 10.10 4.28 9.92 4.15 9.96 4.21
East Urban 3.62 1.41 3.68 1.47 3.64 1.43 3.62 1.41 3.66 1.45
North rural 22.01 10.68 22.37 11.08 22.27 10.91 22.03 10.71 22.15 10.88
North Urban 11.96 5.44 12.06 5.57 12.13 5.57 11.96 5.44 11.92 5.46
West rural 5.35 2.10 5.58 2.31 5.50 2.21 5.38 2.12 5.46 2.20
West Urban 2.35 0.90 2.35 0.90 2.35 0.90 2.35 0.90 2.35 0.90
Total 8.86 3.95 9.04 4.13 9.00 4.05 8.88 3.96 8.93 4.03

From Table 11 and Figure 6, it is clear that households in Northern region were more sensitive to any
scenario in terms of widening the poverty gap than any other region. This was more so pronounced
among the rural households. This was followed by the Eastern region. The Central and Western
regions were not having a big shift in the poverty gaps (P2) for the different scenarios. Indeed,
regions with higher poverty rates tended to show a bigger gap when HIV/AIDS impact was
introduced. The poor households were becoming poorer due to a slight increase in household
expenditure or loss of income.

25

Figure 6: Change in P2 Values as a Result of HIV/AIDS

The different scenarios caused different levels of impact on poverty gaps as summarised in Figure 7.
The scenario that caused such a big gap in poverty was health costs (9%) and reduced incomes due
to AIDS (8.93%) while the funeral costs were the least (8.88%) in widening the poverty gaps. Putting
all these issues together, the poverty gap jumped from 8.86% to 9.04%.
Figure 7: Effects of Different Scenarios on Poverty Gap

Looking at the severity of poverty caused by the different scenarios, from Table 11 and Figure 8, it is
clear that the severity of poverty is increased by the HIV/AIDS epidemic among the households
affected. Individually, health costs (+0.1 percentage points) and lost incomes (+0.08 pp) contribute
greatly to this problem while funeral costs’ effects are minimal (+0.01 pp). In general, all these
factors combined made the severity of poverty jump from 3.95% to 4.13% .

26

Figure 8: Poverty Rates and the Severity of Poverty

Further Understanding of Health Costs
It has been persistently discussed that HIV/AIDS-related health costs are pushing households either
into poverty or deeper into poverty for those that are already poor (as measured by the P0, P1 and
P2 indicators). Because of this, an analysis of the actual composition of health costs in relation to the
total household consumption was computed and illustrated in Figure 9.
Without HIV/AIDS, the health costs at household level were about 10% of all household
consumption expenditure. Factoring in HIV/AIDS, for those households, the health and related costs
would then jump to 45% of household consumption expenditure. This is felt more in rural areas than
in urban areas. Such an increase in the health costs may be compounded by reduced incomes, hence
other necessities of the household are either ignored or reduced. This then pushes the household
into more poverty.
Figure 9: Contribution of Health Costs to total Household Consumption for HIV and no-HIV
Scenarios (%)

27

5. Conclusions
It is evident from the above modelling that HIV/AIDS is likely to increase household poverty levels. In
absolute terms rural households are more affected than the urban households, with a 1.6pp
increase in headcount poverty rates. In proportionate terms, however, urban households are more
affected than rural households.
Regions with high poverty rates tend to experience a smaller impact in terms of increased poverty,
simply because a high proportion of households are already in poverty anyway regardless of
HIV/AIDS. Both proportionately and in absolute terms, it is the better-off regions with lower poverty
rates that experience a larger impact. In general, at the HIV prevalence rate of 6.3%, its impact on
household poverty rate was estimated to be an increase of 1.4pp. The magnitude of the impact of
poverty is comparable with that estimated for Kenya.
The modelling has shown that the major impact on poverty comes from HIV/AIDS health-related
costs (1% increase in headcount poverty), especially in rural households. Other modelled impacts
are smaller, with loss of income due to AIDS contributing 0.5% while funeral costs contributed only
0.1%. However, the impact of health costs on household expenditure is derived from survey work in
South Africa, and may not be accurate in Uganda; hence a more appropriate estimate of poverty
impact could be obtained from a Uganda survey of expenditure patterns in households with HIV+
members. This is both a limitation of this study, and an indication of where further research is
needed.
The long-term impact of HIV/AIDS on household poverty rate was estimated to be smaller than the
short-term impact, at about 0.5%. This is because in the long-term, once household members have
died, the household does not bear additional health costs.
Other than the increase in poverty rate (head count index P0 which shows the number of
households living below the poverty line), the poverty gap (P1) measurements revealed that
HIV/AIDS makes the position of poor households even worse since they are pushed deeper into
poverty. The scenario that had such a big impact as measured by P1 were health costs due to AIDS,
and to some extent loss of income due to the same while the funeral costs had the lowest effect of
pushing households into more poverty. The average poverty gap P1 index for all the scenarios
combined was estimated to be 9.04%, up from 8.86%.
The above effect of HIV/AIDS on poverty was further confirmed by the distributionally sensitive
index P2 by showing the distribution of the income among the poor. Still the health costs and
income adjustments had a higher effect in P2 as opposed to funeral costs. HIV/AIDS made the P2
jump from 3.95% to 4.13%.
While this analysis does not address the impact of ART provision on poverty, the fact that additional
health care costs are the main contributor to increased poverty levels indicates that ART provision
would have a beneficial impact. This is because ART has a significant positive effect on health and
well-being, and will therefore reduce health-related expenditure and increase income levels.
However, this will be offset to the extent that ART provision requires regular visits to health facilities,
which has implications for both household expenditure and time available for work. Further research
evaluating the level of health expenditure in households with HIV+ members and the impact of ART
provision would therefore be worthwhile.

28

Chapter 3: Assessing Sectoral
Vulnerability
1. Introduction
Phase I of the study reviewed the extent to which the impact of HIV/AIDS varied across different
business sectors. This drew upon different types of research, including firm level surveys of HIV-
related impacts and costs, sectoral level modelling (such as Computable General Equilibrium (CGE)
modelling), and sectoral analysis of occupational and demographic structure of the labour force.
A study of firm level impact in South Africa (BER/SABCOHA, 2005) found that the most badly affected
sectors were Mining, followed by Manufacturing and Transport. The impact depends on company
size, skill levels and location. Small and medium enterprises (SMEs) note fewer impacts than medium
and large companies, while companies with predominantly unskilled and semi-skilled workers note a
much greater impact than those employing mainly highly skilled workers.
A range of actual and potential impacts on business are identified below.
• Reduced labour productivity and/or increased absenteeism (especially in Mining, Manufacturing,
Transport and Financial services);
• Higher turnover, recruitment and training costs, and loss of experience and skills of workforce
(Mining, Manufacturing and Transport);
• A smaller impact in retail, wholesale, construction;
• Increased labour demand, including over-staffing in key positions to avoid disruption to
production;
• Some movement towards more capital intensive production techniques (Mining,
Manufacturing);
• Reduced profitability, but little impact on prices.
A second assessment of the sectoral impact of HIV/AIDS in South Africa (USAID/BER 2006) combined
macroeconomic impact analysis with sectoral impact analysis. The latter included an assessment of
sectoral risk, through both the supply side and demand side impact of HIV/AIDS. The supply side
analysis essentially looked at the demographic profile / characteristics of workforces i.e. age, gender
etc., combined with skill structure – to encompass HIV infection risk as well as HIV/AIDS-related
company costs (a given level of prevalence in higher-skilled workforces has a greater cost impact).
This was used to generate a sectoral HIV risk index. On the demand side, the study looked at market
risk, using demand projections from the macroeconomic model. While sectoral HIV infection rates
are not directly measured, company surveys are illustrative, e.g. the prevalence rate in two large
mining companies was 30%, but in four large financial services institutions it was only 3.4%.
The study concluded that high-risk (high prevalence) sectors were Mining, Government,
Manufacturing and Construction, while low risk sectors were Transport, Communications, Business
Services, Finance, and Frade (retail & wholesale). It was noted that although industries with high use
of skilled workers tend to have lower prevalence rates, the cost of infection is higher, given that
skilled workers are more costly to replace and their absence has more impact on production, and
vice versa with semi/unskilled. The study produced a ranking of HIV/AIDS risk by sector, based on
supply side impacts, as shown below.

29

The study also models the channels through which HIV/AIDS affects domestic final demand
(household and government consumption, investment), exports, and intermediate demand (demand
for one industry’s output by another industry). Overall, real GDP growth is lower as a result of
HIV/AIDS, but the effect is not uniform across the economy. Analysis shows that fixed investment is
the component of demand most affected, which feeds through to industries such as construction.
While there is no clear pattern of sectors that are affected on the demand side by HIV/AIDS, it is
primarily those where investment (rather than consumption) demand is important, and where their
output comprises predominantly intermediate demand that is very dependent upon the output of
other sectors. Despite the impact of HIV/AIDS on the population, sectors that are mainly dependent
upon household consumption (community & social services; food & beverages; clothing, etc.) do not
seem to be particularly badly affected.
A study by Rosen et al (2004) reviewed the results of research projects on the company level impact
of HIV/AIDS. Amongst the conclusions of the survey were that a few variables explain most of the
differences in costs among firms. While there was a good deal of variation in costs across and within
countries and sectors, there is also some consistency in the drivers of costs, which are mainly HIV
prevalence in the workforce population; the job level of affected employees (as morbidity and
mortality among more skilled (and higher paid) employees impose higher costs on employers than
they do among less skilled employees); the structure of employment (permanent vs contract and
casual workers); company ownership; and Industrial sector (mining and manufacturing firms face
higher costs than service and agricultural firms, probably as a result of differences in capital
intensity, labour productivity, and workforce demographics).
The present study aims to evaluate the sectoral impact of HIV/AIDS in Uganda through the impact on
the labour force. It uses existing survey data on the occupation and skill structure of different
economic sectors, and the HIV prevalence rate across different sectors and occupations. By sectors,
we refer to the classification of economic activities in the UNHS. It should be noted that the impact
of HIV/AIDS in Uganda may not necessarily be the same as in South Africa, as there are significant
differences in both the economic environment and the HIV/AIDS situation. Compared to South
Africa, the Ugandan economy is much more dominated by agriculture (especially subsistence
agriculture). Uganda also has a much lower HIV prevalence rate, and whereas in South Africa
prevalence rates tend to be inversely-related to skills, education and income (prevalence rates are
higher amongst lower skilled, less educated and less well paid workers), in Uganda the opposite is
true. In South Africa, the high cost of HIV/AIDS stemming from its impact on highly skilled workers
(who are more expensive to replace and train) is partially offset by a lower prevalence rate amongst
such workers, leading to a lower impact in sectors with highly skilled workforces (such as financial
services). In Uganda, however, this may not be the case.

2. Sectoral Impact – HIV Prevalence
The analysis in this chapter is done on the premise that there is no access to ART, and therefore HIV+
people will die as a result of AIDS. The respective sectors of the economy will need to replace them
with new employees to make up for lost labour capacity. This would have a negative economic
impact, due to the additional costs involved. It is then assumed that the impact of these losses
would provide part of the justification for expenditure on ART by government and donors. This is
because it is believed that an HIV+ person that is on ART can live longer, remain healthy, and can
perform to his full capacity in terms of his/her economic participation of the country.

30

This study makes use of the results of the 2004/05 Sero-Prevalence Survey which records HIV
prevalence across a range of individual and household characteristics. The results show that
although the overall adult (15-49 years) prevalence rate was 6.4%, there are considerable variations
around this total by gender, location and wealth status (see Figure 10). In summary, urban HIV
prevalence is higher than rural; there is some evidence of rising HIV prevalence as education
increases; prevalence is higher amongst those working than amongst those not working; there is a
clear positive relationship between HIV prevalence and level of wealth; and female prevalence is
higher than male prevalence.
Figure 10: HIV Prevalence by Residence, Education, Work Status and Wealth

HIV Prevalence by Residence (Urban/Rural) HIV Prevalence by Level of Education

HIV Prevalence by Work Status HIV Prevalence by Wealth Quintile

Source: MoH, 2006 (Sero survey)

The survey also provides data on sero-prevalence across economic sectors (see Figure 11). This
shows major variations in the HIV prevalence rate between sectors. By far the highest is Public

31

Administration, with a prevalence rate of 16.3%. Agriculture has a relatively low prevalence rate of
6.0%, with prevalence being higher in most non-agricultural sectors of the economy11.
Figure 11: HIV Prevalence by Sector

Source: Calculations based on Sero survey data (MoH, 2006)

Although the prevalence rate is (relatively) low in the Agricultural sector, it is of course by far the
largest sector of the economy. Hence if we consider the estimated numbers of HIV+ workers across
the economy, a different picture emerges (Figure 12).

11
Note that the sample size was very small in mining, financial intermediation, extra-territorial organisations
and electricity, gas & water, and these sectors are grouped together as “other”.

32

Figure 12: Distribution of HIV+ Workers by Sector


Looking at HIV prevalence by sector and by gender, Figure 13 shows that a higher proportion of
female employees are likely to be HIV+ than their counterpart male employees. However, there are
a few sectors such as Construction, Transport and Communications that show a contrasting picture,
with a higher male prevalence rate. In Public Administration, the difference in the male and female
prevalence rates is small.

33

Figure 13: HIV Prevalence by Gender and Sector


With regard to the absolute numbers of HIV+ workers in each sector, the picture may differ from
that of prevalence rates, since some sectors may be dominated by men and others by women. The
distribution of male and female HIV+ workers in each sector is shown in Figure 14. Therefore, in
Construction, Transport & Communications, Fishing, Mining and Public Administration, where the
workforce comprises predominantly males, the latter account for more than 75% of the HIV+
workforce in each sector. By contrast, in domestic employment and hotels & restaurants, females
make up the majority of the workforce, and more than 80% of the HIV+ workers in these sectors.
Similarly in agriculture, two-thirds of the HIV+ workers are women.

34

Figure 14: Distribution of HIV+ Workers by Sector and Gender


3. Sectoral Impact – Occupational Structure
The sero survey also provides information on HIV prevalence by occupation. As noted earlier, there
is a higher prevalence rate amongst working adults than non-working adults. Amongst working
adults, there is some evidence that HIV prevalence varies across occupations (Figure 15). The
prevalence rate is relatively high amongst Sales, Clerical and Service sector workers, who might
generally be classed as semi-skilled. There is a slightly lower, but above average, prevalence rate for
skilled professional and manual workers. The lowest prevalence rates are for the unskilled categories
of manual and agricultural workers. These results confirm that unskilled workers have lower HIV
prevalence rates than semi-skilled and skilled workers which is contrary to the findings in South
Africa.
Figure 15: HIV Prevalence by Occupation

35


The information on HIV prevalence by occupation can be combined with information on the sectoral
composition of the labour force in different sectors to further analyse sectoral vulnerabilities. First
we can calculate a “proxy” HIV prevalence rate by sector, taking into account these two pieces of
information. This is shown in Figure 16 below, and is calculated by combining occupational HIV
prevalence rates with the occupational structure of the various sectors. The results are largely
consistent with the earlier results (in Figure 11), although the prominence of Public Administration is
reduced. We also have a proxy prevalence rate for financial intermediation, which suggests that it is
one of the most vulnerable sectors.
Figure 16: “Proxy” HIV Prevalence Rates by Sector (derived from occupations)


36

The above data take account only of the occupational composition of different sectors and
occupational prevalence rates. We can also incorporate the varying importance of different labour
occupations to the output of each sector, i.e., reflecting the fact that a skilled worker makes a
greater contribution to output than an unskilled worker, and is more difficult to replace. Hence,
losing a skilled worker is likely to be more disruptive to production than losing an unskilled worker.
While the direct contribution of different labour categories cannot be directly measured, we can
approximate this from the relative wage rates, which are available from the 2005/06 UNHS.
Table 12: Median Monthly Wages by Occupation

Occupation Monthly wages Differential (relative to
(Shs ‘000, median) unskilled elementary
occupations)
Legislators, Managers 120 5.0
Professionals 250 10.4
Technicians & Associate Professionals 148 6.2
Clerks 80 3.3
Sales & Service 50 2.1
Agriculture & Fisheries 27 1.1
Crafts & related Trades 91 3.8
Plant & Machine Operators 91 3.8
Elementary Occupations 24 1.0
Source: UNHS 2005/06, Table 4.8 & own calculations

Taking this into account, the relative sectoral vulnerabilities are shown in Figure 17. These primarily
reflect each sector’s dependence on skilled labour, as well as the variations in HIV prevalence across
occupational categories. The most vulnerable sectors are Education, Health and Social Work, Finance
as well as Public Administration, because of their high dependence upon skilled workers.
Figure 17: Index of Sectoral Vulnerability

Source: Own calculations

37

An alternative approach to evaluating sectoral vulnerability is to consider the cost of educating and
training workers at different levels. To the extent that HIV+ workers will die in the absence of
treatment and would have to be replaced, then the cost of educating and training these workers
represents a burden on the economy.
Information on the level of education of workers in different sectors is available from the 2006
Labour Market Conditions Report. The overall composition of the labour force by education level is
shown in Table 13.
Table 13: Education Level by Sector (% of workforce in sector)

No Some Comple- Some Comple- Post Do not
formal primary ted P7 secondary ted S6 secondary know
educ.
Agriculture etc. 17.5% 51.4% 13.6% 16.3% 0.6% 0.5% 0.1%
Fishing 13.8% 48.6% 17.4% 17.4% 0.9% 1.8% 0.0%
Mining 4.5% 40.9% 27.3% 18.2% 4.5% 0.0% 4.5%
Manufacturing 10.0% 40.4% 17.0% 25.8% 2.0% 4.6% 0.2%
Electricity, Gas and Water 0.0% 30.0% 0.0% 20.0% 20.0% 30.0% 0.0%
Construction 5.0% 30.8% 21.4% 33.8% 2.0% 7.0% 0.0%
Sales 8.6% 37.9% 18.4% 28.7% 1.9% 4.4% 0.1%
Hotels and Restaurants 15.1% 36.2% 17.9% 25.2% 0.5% 4.6% 0.5%
Transport, Storage and 1.8% 36.0% 21.3% 36.4% 1.1% 2.9% 0.4%
Comms.
Financial Intermediation 0.0% 0.0% 0.0% 17.6% 11.8% 70.6% 0.0%
Real Estate, Renting and 0.0% 16.0% 4.0% 54.0% 8.0% 18.0% 0.0%
Business
Public Administration 1.0% 15.3% 8.2% 35.7% 8.2% 31.6% 0.0%
Education 1.1% 2.6% 2.3% 46.0% 4.3% 43.7% 0.0%
Health and Social Work 3.5% 9.2% 8.5% 44.0% 1.4% 33.3% 0.0%
Other Community, Social 8.8% 34.0% 14.4% 32.6% 4.7% 5.1% 0.5%
and Pers. Activities
Domestic Service 19.5% 50.0% 15.3% 14.4% 0.0% 0.0% 0.8%
Extra Territorial 0.0% 22.2% 0.0% 55.6% 11.1% 11.1% 0.0%
Organisations
Total 14.7% 46.1% 14.1% 20.4% 1.1% 3.4% 0.1%

Source: Calculations based on UNHS data (UBOS, 2007)

This information can be combined with estimates of the cost of educating students to different
levels, as shown in Table 14.
Table 14: Cost of Education

Level Cost per Total cost to this
term/semester level (Shs)
(Shs)
No formal education 0 0
Primary years 1-4 20 000 240 000
Primary years 5-7 50 000 690 000
Secondary years 1-4 150 000 2 490 000
Secondary years 5-6 200 000 3 690 000
Post secondary 1 200 000 10 890 000

38

This information can be used to calculate the total cost of educating workers in the sector who may
die due to HIV/AIDS (see Figure 18). The cost depends on the size of the sector (number of workers),
the level of education of workers in the sector, and the sectoral HIV prevalence rate. While
agriculture faces the biggest cost (34% of the total cost), it is proportionately much less than its
share of employment (74%), because of the relatively low level of education of workers in the sector.
The education sector faces a relatively high cost (16%) relative to its employment share (3%), due to
its high dependence upon educated workers.
Figure 18: Total Cost of Educating HIV+ Workers (% by sector)


A more accurate assessment of the cost burden can be obtained by considering the average cost of
educating a worker in the sector (see Figure 19).

39

Figure 19: Cost of Education of an Average Worker, by Sector (Shs million)


This cost can be further related to the wage bill in the sector, i.e., the cost of replacing HIV+ workers
as a proportion of the annual wage bill. On the assumption that HIV+ workers live on average ten
years after getting infected, the burden is shown in Figure 20. The greatest burden is in the Hotels &
Restaurants sector, which mainly represents the impact of relatively low wages in the sector
combined with a high HIV prevalence rate. The next highest burdens are in Education, Public
Administration and Health & Social Work, reflecting the high level of education of workers in these
sectors and relatively high prevalence rates. While the Financial Services sector has both a relatively
high prevalence rate and highly educated workers, it also has the highest wage levels of any sector
so the relative cost of educating workers is reduced.

40

Figure 20: Cost of Replacing HIV+ Workers (as % of annual wage bill)


4. Cost of Providing Anti-Retroviral Therapy (ART)
Using the above information, we can compare the cost of education of workers in different sectors
with the cost of providing ART. Education costs across sectors are shown in Figure 19 above, and
varies from Shs 590 000 in domestic service to Shs 8 561 000 in financial intermediation. In terms of
the USA dollar, the cost of education varies from approximately US$ 350 to US$ 5 000. This
compares with the cost of providing ART, which is around US$ 500 a year per person. If we assume
that the average (median) additional life-years resulting from ART is ten, then the total cost of
providing ART to an HIV+ person is US$ 5000. In straightforward financial terms, therefore, ART is
not a good investment when compared to the cost of educating workers, as the cost of the lost
education is less than the cost of providing ART in all sectors except for the Finance sector. While the
total cost of providing trained, experienced and educated workers is greater than the cost of
education used here (the study is limited by lack of information on the costs of on-the-job training
and the value of experience), this example nevertheless illustrates that ART may not be a good
investment in purely financial terms, at least not for all workers.
An alternative perspective on this issue can be gained by considering the cost of ART relative to
wage costs; if we assume that the value of a worker’s contribution to output is approximately equal
to the wage he or she receives, then this provides a crude comparison of the costs of ART with the
value of output provided by a worker who continues to contribute to produce. Wage costs by sector
are shown in Figure 21. Only in a few sectors (Financial Services; Electricity, Gas & Water; as well as
Public Administration) are average wages significantly above the annual US$ 500 cost of ART,
although in several other sectors, average wages are just above US$ 500. This again suggests that
the cost of providing ART may not be justified in financial terms in many sectors of the economy.

41

Figure 21: Average Wage by Sector (US$/year)


5. Summary and Conclusion
This study presents an analysis of the impact of HIV/AIDS on different economic sectors, measured
by the impact on the labour force in each sector. This impact was assessed primarily through the
level of education of workers in each sector, and the cost of educating replacement workers to make
up for those who die as a result of HIV/AIDS. These costs therefore depend on the level of education
of workers in each sector, and differences in HIV prevalence by level of education and across sectors.
The study does have some limitations. First, the costs relate specifically to educating workers at
different levels and do not take account of the loss of experience gained while working. Therefore
the costs of replacing workers in the Agricultural sector may be under-estimated, given that they
generally have low levels of education but significant skills acquired through experience. The second
limitation is that the costs of recruitment and on-the-job training could not be quantified and are
therefore omitted from the analysis. This could lead to an under-estimate of the cost of replacing
people that die of HIV/AIDS, especially in sectors such as Financial Services and Public Administration
where there are high levels of job-specific training, and potentially lengthy recruitment processes,
especially for senior staff.
Subject to these limitations, the overall results of the sector vulnerability analysis can be
summarised as follows:
• The HIV prevalence rate (measured directly) is highest in Public Administration, followed by
Hotels & Restaurants, Sales, and Fishing. When measured indirectly, the prevalence rate is
highest in Hotels & Restaurants, followed by Sales, Finance and Public Administration.
• The total number of HIV+ workers is highest in Agriculture, followed by Sales.

42

• The cost of educating a worker is highest in Financial Intermediation (at P8.6 million Shs),
followed by Education (at Shs 6.1 million), Health (at Shs 4.9 million) and Public
Administration (at Shs 4.7 million).
• The total cost of replacing HIV+ workers (through the investment needed in their education)
is highest in Agriculture (34% of the total), followed by Education (19%) and Sales (16%).
• The cost burden of replacing an HIV+ worker, as a proportion of the wage bill, is highest in
Hotels & Restaurants, Education and Public Administration.
• Relative to the costs of educating workers, or average wages (as an approximation of the
value of output produced), providing ART may not be a good investment in purely financial
terms.

43

6. Appendix: Data Tables
Sector HIV prev Total labour Av wage Total Wage Bill No. HIV+ Total Education Cost per Cost per Total Annual
rate force (2002 (UGX/month) (UGX cost and training HIV+ worker (all cost as cost as %
(proxy) Census & million/month) to replace those worker that workers) % of of
Labour Mkt dying due to AIDS may die annual annual
survey) wage bill wage bill
Agriculture, etc. 5.3% 7,828,900 22,000 172,236 409,037 286,625 700,730 36,611 14% 1.4%
Fishing 7.1% 105,560 77,000 8,128 6,661 6,026 904,679 57,088 6% 0.6%
Mining 6.8% 14,849 22,000 327 1,295 1,215 938,182 81,841 31% 3.1%
Manufacturing 7.4% 442,895 60,000 26,574 32,076 45,959 1,432,800 103,769 14% 1.4%
Elec, Gas & Water 7.9% 7,296 160,000 1,167 683 3,127 4,575,000 428,564 22% 2.2%
Construction 7.1% 147,342 75,000 11,051 12,395 23,500 1,895,970 159,496 18% 1.8%
Sales 9.8% 869,193 60,000 52,152 88,537 131,318 1,483,206 151,080 21% 2.1%
Hotels and Restaurants 10.3% 170,137 33,000 5,615 19,345 26,275 1,358,257 154,433 39% 3.9%
Transport 7.0% 210,103 80,000 16,808 16,358 24,594 1,503,419 117,055 12% 1.2%
Communications
Financial Intermediation 9.2% 16,818 500,000 8,409 1,352 11,576 8,560,588 688,296 11% 1.1%
Real Estate, Renting and 8.6% 46,369 100,000 4,637 3,702 13,573 3,666,000 292,720 24% 2.4%
Business
Public Administration 9.2% 71,582 140,000 10,021 7,761 36,697 ,728,367 512,650 31% 3.1%
Education 8.6% 281,303 100,000 28,130 26,106 158,883 6,085,971 564,811 47% 4.7%
Health and Social Work 8.5% 102,073 100,000 10,207 10,366 50,359 4,858,085 493,362 41% 4.1%
Other Comm, Soc and 8.1% 179,172 20,000 3,583 15,070 25,976 1,723,674 144,979 60% 6.0%
Pers Activities
Domestic employment 6.9% 91,352 20,000 1,827 7,038 4,151 589,831 45,440 19% 1.9%
ETOs 8.0% 4,042 100,000 404 622 1,900 3,056,667 470,002 39% 3.9%
Total 6.2% 10,588,986 361,276 653,472 736,222 1,126,630 17% 1.7%

44

Chapter 4: HIV Costing, Financing and
Expenditure
1. Introduction
One of the major determinants of the macroeconomic impact of HIV/AIDS expenditure is the extent
to which that spending is sourced domestically (from the Government budget) or externally (from
donor funds). A second important aspect is whether the funds are spent externally (on, for instance
imported drugs) or domestically (e.g., local salaries or purchase of locally-produced inputs). Most of
the concerns about the impact of HIV/AIDS on expenditure stems from concerns that large amounts
of external funds flowing into the country boost aggregate demand, which in turn causes inflation
and real exchange rate appreciation (and loss of international competitiveness), and destabilise the
macroeconomic achievements that have been secured over the past 15 years. However, this effect is
reduced when the greater proportion of spending is devoted to imported goods and services, as
domestic aggregate demand is less affected.
While there is some information on the sourcing of HIV/AIDS funding, there is little or no information
on how the money is spent, or what it is spent on. The objective of this assignment is to track the
flow of resources received through to spending, to determine what HIV/AIDS-related funds are spent
on, and in particular, whether that expenditure is on domestic or imported goods and services.
The overall flow of funding for HIV/AIDS-related activities is depicted in Figure 22, at four levels: (i)
Sources of finance; (ii) Spending agencies; (iii) Projects; and (iv) Spending items.
Figure 22: HIV/AIDS Funding and Expenditure

45

High level data on overall HIV/AIDS spending and financing from 2003/04 to 2006/07 were collected
by Lake & Mwijuka (2006) (see Figure 23). This focused on overall spending levels and sources of
financing, and illustrated the rapid increase in overall spending on HIV/AIDS programmes over this
period, rising from US$ 38 million to US$ 164 million in four years. As the chart illustrates, this was
largely driven by greatly increased funding from the USA. Almost all spending was financed externally
– domestic funding from the GoU increased from US$ 6 million to US$ 8 million over this period, but
at the same time the GoU’s share of total spending fell from 16% to 5%.
Figure 23: Financing of HIV/AIDS-related Spending, 2003/4 - 2006/7

Source: Lake & Mwijuka (2006)

2. Methodology
The focus of this study is not on overall funding levels but on providing more detail on spending. The
methodology followed has been to gather information from resource providers (donors and the GoU)
relating to the sources of funds, and from entities involved in spending those resources. Information
has been sought on the main categories of expenditure, and on whether that expenditure was mainly
domestic or external. The emphasis has been on tracking the main financial flows, rather than all
financial flows. Owing to the nature of the exercise, this information seems incomplete, but the
objective is to isolate the main flows in order to ascertain the macroeconomic magnitudes.
The categories of spending identified were:
• Salaries
• Allowances
• Technical Assistance
• Drugs & Medical Services
• Information, Education and Communication (IEC)
• Monitoring & Evaluation (M&E)

46

• Training
• Transport/vehicles
• Miscellaneous Supplies
• Other
An instrument was developed to collect the necessary data from donors and implementing agents,
and results were obtained from a range of financing and implementing agencies (see Appendix).
Information was requested for the years 2004/05, 2005/06 and 2006/07. Spending entities were
identified by the President’s Emergency Plan for AIDS Relief (PEPFAR) list of Prime Partners (see
Appendix) and other sources. PEPFAR is by far the dominant funder of HIV/AIDS-related activities in
Uganda, with budgeted spending increasing from US$ 135 million in FY2005 to US$ 236 million in
FY2007.

3. Results
Unfortunately, it was not possible to get the data in as consistent a form as desirable, due to
variations in the quality of information across agencies. In particular, many agencies did not keep
information in a form which enables the disaggregation of spending into the categories required by
the project, especially with regard to domestic versus external spending, while others were unwilling
to provide such information even where it existed.
In the public sector, expenditure on HIV/AIDS was not well distinguished from expenditure on other
health activities. For instance, expenditure on blood transfusion services to ensure provision of safe
blood to patients includes expenditure on ensuring that blood is free from HIV but also free from
hepatitis B, malaria, syphilis, etc. It is therefore difficult to know what proportion of the money is
spent on HIV at blood transfusion services. The same situation applies to other goods and services
e.g. human resources, vehicles, infrastructure, etc.
Information on HIV/AIDS-related spending over the period 2004/05 to 2006/07 was obtained from
the following institutions (see Table 15), covering two-thirds (estimated at 67%) of total spending
over the relevant period.

47

Table 15: HIV/AIDS-related Spending by Institution, 2004/05 – 2006/07 (US$)

Institution 2004/5 2005/6 2006/7 Total
Christian Children's Fund (CCF) 421,882 614,878 528,515 1,565,275
AVSI 226,197 677,378 955,790 1,859,365
Elisabeth Glaser Pediatric AIDS Foundation 2,100,000 3,125,000 4,220,000 9,445,000
Protecting Families Against HIV/AIDS (PREFA) 1,299,584 1,532,477 2,832,061
Uganda Blood Transfusion Service (UBTS) 1,715,659 2,660,689 4,010,411 8,386,760
Population Services International (PSI) 3,275,511 2,812,718 3,635,755 9,723,985
International HIV/AIDS Alliance 709,877 709,877
The AIDS Support Organisation (TASO) 10,901,546 16,598,357 18,870,191 46,370,095
Global Fund Against AIDS, Tuberculosis and 10,951,458 10,951,458
Malaria (GFATM)
Mildmay 3,804,083 6,095,423 9,899,506
Uganda AIDS Commission 17,391,831 12,653,389 9,515,461 39,560,681
Uganda Virus Research Institute 241,554 170,000 144,809 556,363
MoH (ACP/STD project) - 653,700 248,373 902,073
International Youth Federation - 321,845 248,690 570,535
National Medical Stores 11,408,150 33,306,262 12,410,028 57,124,440
Joint Clinical Research Centre 6,430,580 8,495,586 11,510,289 26,436,454
Catholic Youth Services - - 13,422,117 13,422,117
Baylor University - - 1,942,568 1,942,568
Mulago-Mbarara - 4,789,523 5,824,528 10,614,051
Plan 1,070,426 2,471,079 1,828,093 5,369,598
Care 3,475,745 8,442,481 10,649,241 22,567,467
Total 69,610,540 102,896,552 108,302,637 280,809,729
Estimated total expenditure, 2004/5 - 2006/7 103,331,410 150,542,306 164,396,347 418,270,063
Proportion covered 67.4% 68.4% 65.9% 67.1%
Source: Own calculations from study data

The composition of this spending is shown in

48

Figure 24.

Figure 24: Breakdown of HIV/AIDS Spending 2004/5 - 2006/7


Separate, but less detailed data have been obtained regarding PEPFAR funds, which are by far the
largest source of funding for HIV/AIDS-related spending. The breakdown of PEPFAR spending over
the period 2005/07 is shown in Figure 25.

49

Figure 25: Breakdown of PEPFAR Spending, 2005-7

Source: Own calculations from PEPFAR data

The quality of information regarding the breakdown of expenditure between domestic and external
spending has not been up to expectation. However, applying estimated proportions of external
spending to the various categories of expenditure (ranging from 95% of treatment costs, which are
primarily imported drugs, to 10% for information, education and communication (IEC) and training,
which are mostly carried out by local staff), enables an estimate of the external spending. As a result,
it is estimated that almost 60% of total spending goes (directly or indirectly) on external goods and
services, while 40% is spent domestically. The high proportion of donor receipts spent externally
significantly reduces the potential macroeconomic impact of aid inflows12.
Table 16: External Component of HIV/AIDS-related Spending (2004/5 – 2006/7, US$)

Category External
Salaries & Allowances 39,083,272 33% 13,027,757
TA/Consultants 13,392,253 50% 6,696,127
Treatment (incl. ART) 114,963,905 95% 109,215,710
Supplies 18,658,002 67% 12,438,668
IEC, Counselling, etc. 27,869,863 10% 2,786,986
M&E 12,242,485 10% 1,224,248
Training 14,328,063 10% 1,432,806
Vehicles, Transport & Travel 8,097,254 50% 4,048,627
Other 29,381,760 33% 9,793,920
Total 278,016,857 58% 160,664,850

Carrying out a similar process for PEPFAR funds yields a similar total (56% external spending).

12
These estimates include both the direct and indirect external spending components.

50

Table 17: External Component of PEPFAR Spending (FY2005-2007, US$)

Amount External % External Amount
Prevention 85 214 972 25% 21 303 743
Care 159 597 138 25% 39 899 285
Treatment 205 697 121 95% 195 412 265
Other 73 225 946 50% 36 612 973
Total Funding 523 735 177 56% 293 228 265
Source: Own calculations from PEPFAR data

4. Future Resource Needs and Funding
Estimates of future resource needs for HIV/AIDS programmes are provided in the UAC’s NSP for
2007/08 to 2011/12. Based on the “high funding” scenario, projected spending will approximately
triple from US$ 170 million in 2006/07 to US$ 511 million in 2011/1213.
The NSP projections envisage that 85% of the funding would come from donors and 15% from the
GoU. This represents an increase in the GoU’s share from 5% in 2006/07. In view of the anticipated
increased share of total spending to be met by the GoU, and the increase in total spending, these
projections entail an increase in GoU funding from US$ 8.2 million in 2006/07 to US$ 75 million in
2011/12, a nine-fold increase. Donor funding would increase from US$ 162 million to US$ 436 million
over the same period, nearly a three-fold increase. While the anticipated increased GoU funding
represents a sharp rise in the proportion of domestic revenues devoted to HIV/AIDS, it should still be
sustainable14. However, if donor funds were not available, then it is unlikely that the same level of
programmes could be sustained on the basis of domestic resources, and that significant cutbacks
would be necessary. This is because an entirely domestically-funded programme would consume an
unsustainable proportion of projected domestic revenues, requiring either a sharp increase in
taxation or a large budget deficit, both of which would have highly negative economic impacts that
would reduce economic growth.

5. Conclusion
Although the response to requests for data was disappointing, the study has nonetheless yielded
some useful results. Specifically, we have derived some plausible, albeit uncertain, estimates of the
split between domestic and foreign spending in HIV/AIDS programmes. The conclusion that
approximately 60% of total spending is devoted to imported goods and services indicates that the
net macroeconomic impact on the balance of payments (BoP), exchange rate, money supply, etc. is
considerably less than the gross impact. If the same proportion applies to total spending (and the
fact that data was received on two-thirds of spending suggests that the aggregate figure would not

13
The “high funding” scenario assumes that generous donor support will continue to be provided, enabling a
substantial scale-up from present levels of programme provision. However, it is below the “full funding”
scenario that would enable full coverage of every intervention, and which would cost US$ 680 million by
2011/12; hence the envisaged scenario entails scaling down of some targets. For further details, see Section
4 of the NSP (Resource Requirements).
14
The budgetary and macroeconomic impacts of GoU spending on HIV/AIDS are discussed in more detail in
the Phase III report.

51

be much different) this then means that of the total estimated spending for HIV/AIDS programmes of
US$ 418 million over the period 2003/04 to 2006/07, some US$ 243 million was spent on externally-
sourced goods and services, while an estimated US$ 176 million was spent domestically.
Going forward, the proportion of total expenditure that will be spent externally is likely to increase,
given that spending on ART drugs is set to rise sharply under the NSP 2007/08 – 2011/12, and this
has the highest import component of any HIV/AIDS programme. However, to the extent that ART
drugs are produced locally, the import content would be reduced, and this would tend to worsen the
adverse macroeconomic impacts.
Given the concerns expressed by policymakers that the inflow of donor funding for HIV/AIDS
programmes may cause macroeconomic disturbance, specifically by causing the exchange rate to
appreciate or potential inflationary pressures, these results show that any such adverse
developments would be substantially less than that suggested by the “headline” spending numbers.
Furthermore, it is important that analysis of the macroeconomic impact of HIV/AIDS inflows takes
account of the offsetting effect of external purchases of goods and services.

52

Appendix 1 – Data Collection Form

Implementing
agent
Source of
funding
Amounts
received
Year/period
Expenditure
Local
Salaries Foreign
Total
Local
Allowances Foreign
Total
Local
Technical
Foreign
assistance
Total
Local
Drugs Foreign
Total
Local
Supplies Foreign
Total
Local
IEC Material Foreign
Total
Local
Monitoring &
Foreign
Evaluation
Total
Local
Training Foreign
Total
Local
Vehicles Foreign
Total
Local
Other (specify) Foreign
Total

53

Appendix 2: PEPFAR Uganda Partners: FY 2006

Institution Funding (US$)
African Medical and Research Foundation 1,550,000
AIDS Information Centre 3,981,119
Baylor University, College of Medicine 2,737,252
Catholic Relief Services 8,784,303
Chemonics International 1,400,000
Emerging Markets 700,000
Family Health International 300,000
HOSPICE AFRICA, Uganda 900,000
Integrated Community Based Initiatives 905,000
International HIV/AIDS Alliance 1,100,000
International Medical Corps 225,000
International Rescue Committee 375,000
John Snow, Inc. 9,737,851
Johns Hopkins University Center for Communication Programmes 4,150,000
Johns Hopkins University Institute for International Programmes 200,000
Joint Clinical Research Center, Uganda 13,422,060
Kumi Director of District Health Services 795,000
Makerere University Faculty of Medicine 6,520,365
Makerere University Institute of Public Health 2,235,870
Medical Research Council of Uganda 600,000
Mildmay International 7,994,682
Ministry of Health, Uganda 2,575,000
National Medical Stores 3,900,000
New York AIDS Institute 300,000
Population Services International 3,004,929
Protecting Families Against AIDS 1,115,076
Research Triangle International 466,000
Social and Scientific Systems 1,350,000
The AIDS Support Organisation 16,863,700
Uganda Virus Research Institute 170,000
University of California at San Francisco 880,000
US Agency for International Development 3,178,818
US Centers for Disease Control and Prevention 10,613,639
US Department of Defense 225,000
US Department of State 315,734
US Peace Corps 600,700
Walter Reed 714,400

Source: PEPFAR website

54

Appendix 3: Responses Received

Organisation Remarks
African Medical Research Foundation (AMREF) Data required were received.
Africare No data obtained yet.
AIDS Information Centre Data required were received.
AVSI Data required were received.
Baylor University Data required were received, for 2006/07 only
Catholic Relief Services Data required were received, for 2006/07 only
Christian Children’s Fund Data required were received.
Elizabeth Glaser Paediatric AIDS Foundation Data required were received.
Family Health International - Uganda No data obtained yet.
Health Communications Partnership No data obtained yet.
Hospice Africa-Uganda No data obtained yet.
IAVI – Entebbe No data obtained yet.
International HIV/AIDS Alliance Data required were received, for 2006/07 only
International religious council of Uganda Data required were received.
International Youth Foundation Data required were received, except for 2004/05
Joint Clinical Research Centre Data required were received.
Makerere University Rakai Project No data provided
Makerere University Walter Reed Project Data required were received
Mbarara-Mulago Joint AIDS Programme Data required were received, except for 2004/05
Medical Research Council No data provided
Mildmay Data required were received, except for 2004/05
Ministry of Finance Data required were received
Ministry of Health (ACP) Data required were received, except for 2004/05
Ministry of Local Government Data required were received
National Medical Stores Data required were received.
PEPFAR National Coordinator Some data received, but with limited detail
Population Service International Data required were received
Protecting Families Against HIV/AIDS Data required were received, except for 2004/05
Research Triangle International Data required were received
The AIDS Support Organisation (TASO) Data required were received.
Uganda AIDS Commission Data required were received.
Uganda Blood Transfusion Services Data required were received.
Uganda Global Fund AIDS, Tuberculosis and Data required were received.
Malaria Project (UGFATMP)
Uganda Virus Research Institute Data required were received.
UNAIDS Data available.
WHO Data available.
World Vision International No data provided

55

Chapter 5: The Demographic Impact
of HIV/AIDS in Uganda
1. Introduction
The preparation of demographic projections is an essential component of modelling the
macroeconomic impact of HIV/AIDS. A significant component of the macroeconomic impact results
from the effect of HIV/AIDS on the population, and hence on the size and growth of the labour force.
The labour force is in turn one of the main long-term drivers of economic growth, and also impacts
on relevant indicators such as wages, employment, and the relative growth of different economic
sectors.
The Spectrum model15 was used to prepare demographic projections for this study. Spectrum has a
number of advantages for this purpose, including its ease of use, and the relatively limited range of
data that is needed to calibrate the model. It also has a module dedicated specifically to modelling
the impact of HIV/AIDS (the AIDS Impact Module – AIM), and can produce a range of relevant
outputs relating to the impact of HIV/AIDS on the population. It can also accommodate treatment
interventions, such as the provision of ART.
Spectrum is widely used to make projections of population and resource needs in the context of
HIV/AIDS. Amongst others, the Spectrum AIM is used by the Joint United Nations Programme on
HIV/AIDS (UNAIDS) to make the national and regional estimates which are released once every two
years. In Uganda, the Uganda Bureau of Statistics (UBOS) has prepared national population
projections using the same model.
Spectrum requires the following inputs in order to provide basic demographic projections:
• Base year population (pre-HIV/AIDS).
• Life expectancy trends for males and females (in the absence of HIV/AIDS).
• Total fertility rate.
• Age specific fertility rate.
• Sex ratio at birth.
• International migration.
• Model life table
Spectrum provides default parameters for all of the above. However, in some cases the default
parameters were replaced with actual Uganda data.
This study describes the process used to prepare the projections using Spectrum, and presents
selected results.

2. Scenarios
Projections were prepared for four different scenarios:

15
The Spectrum model is freely available at www.constellagroup.com/international-
development/resources/software.php. Further details on Spectrum are provided at Appendix 1.

56

1. No AIDS.
2. AIDS with no ART provision.
3. AIDS with “low” ART provision.
4. AIDS with “high” ART provision.
The “No AIDS” scenario provides hypothetical population projections for Uganda in the absence of
HIV/AIDS.
The “AIDS with No ART” scenario introduces HIV/AIDS into the projections, but does not include the
impact of any treatment interventions.
The “AIDS with low ART” scenario also includes the impact of HIV/AIDS on the projections, but also
includes the impact of the introduction of ART. It assumes that ART provision remains at relatively
low levels.
The “AIDS with High ART” scenario also includes the impact of HIV/AIDS and ART provision on the
projections. However, it assumes that ART provision continues to grow steadily from current levels.
All scenarios include population projections through to 2025.

3. Population Projections
Basic Demographic Assumptions
The following assumptions were made to project the population to 2025.
The Base Population: The base population (year 0 of the projections) should ideally relate to a year
prior to the beginning of the HIV/AIDS epidemic. Otherwise, it is not possible to derive a “No AIDS”
scenario, because the impact of HIV/AIDS will already be factored into the base year population. This
was somewhat problematic due to the limited availability of detailed historical census data, given
that the base year population would have to be 1980 or earlier. The projections therefore used the
“Easyproj” module of Spectrum, which uses data from UN Population Division. The initial population
for 1980 taken from the projections is shown in Table 18.

57

Table 18: Population in 1980 by Age & Gender

Age group Male Female Total
0-4 1,256,700 1,242,000 2,498,700
5-9 999,100 989,600 1,988,700
10-14 814,100 809,700 1,623,800
15-19 642,100 641,800 1,283,900
20-24 522,700 527,500 1,050,200
25-29 429,000 437,800 866,800
30-34 357,300 362,400 719,700
35-39 292,800 296,700 589,500
40-44 217,500 221,900 439,400
45-49 200,700 206,100 406,800
50-54 162,900 169,900 332,800
55-59 128,400 136,800 265,200
60-64 97,000 103,900 200,900
65-69 67,600 74,500 142,100
70-74 43,400 49,700 93,100
75-79 24,000 29,400 53,400
80+ 13,400 18,500 31,900
Total 6,268,700 6,318,200 12,586,900
Source: Spectrum model

Sex Ratio at Birth: Spectrum requires information on the sex ratio of the population at birth.
Whereas it is appreciated that vital registration provides the most appropriate source of information
on sex ratio at birth, the coverage of vital registration in Uganda is still very limited. From the UDHS
results, the sex ratio at birth was estimated at 102.6 males per 100 females since 1980 and this was
assumed constant throughout the projection period.
Net Migration: Spectrum uses data migration from UN sources. Estimates are shown in Figure 26
below. Flows of migrants are determined largely by changing political and humanitarian
circumstances in regional states. Historical data are used where available. Going forward, projected
flows are relatively small.

58

Figure 26: International Migration


Fertility: It was assumed that total fertility rates (TFR) will continue to decline as noted between the
Censuses of 1991 (7.0) and 2002 (6.7). If this trend continues, it is assumed that by 2025, the TFR will
then be 5.9. The TFR in 1980 was assumed to be 7.1.
Life Expectancy: Spectrum requires life expectancy figures in the absence of HIV/AIDS. The figures
used were those generated by the model and are shown in Figure 27 below.

59

Figure 27: Life Expectancy, Male & Female


HIV/AIDS Projections
The AIM of Spectrum requires information on various parameters associated with HIV/AIDS. These
include the adult HIV prevalence rate, HIV age distribution, the fertility impact of HIV/AIDS, the
survival time after infection, and the extent of interventions. The assumptions and parameters are
discussed below.
Adult HIV/AIDS Prevalence: Estimates of HIV/AIDS prevalence rates for Uganda are available from
different sources. However, prior to the advent of sentinel survey testing of pregnant women,
estimates are of uncertain reliability with regard to national prevalence rates, and tend to reflect
prevalence in specific localities. Uganda AIDS Commission (2001) quotes adult HIV prevalence rates
of 9% for 1988, rising to 18.5% in 1995, and falling to 8.3% in 199916. As per the Sero-Behavioural
survey of 2004/05, the prevalence was estimated at 6.4%. Following this trend, the prevalence was
projected to be 4.5% in 2025.
Plotting the publicly available prevalence rates shows a series that does not appear to represent a
normal epidemiological trend, with discontinuities that do not occur in practice. This may reflect the
fact that data on HIV prevalence were taken from a variety of sources. Therefore, in order to
generate usable HIV prevalence rates, the series was smoothed17. The smoothed series shows a
lower peak in HIV prevalence and a more gradual decline than the original data. Both the original
and smoothed series are shown in Figure 28 below.

16
Twenty Years of HIV/AIDS in the World: Evolution of the Epidemic and Response in Uganda (UAC, 2001).
17 th
The smoothing was applied by fitting a 4 -order polynomial function in Excel.

60

Figure 28: Adult HIV Prevalence

Source: UAC, UBOS

Age- and gender-specific adult HIV prevalence rates were taken from the results of the 2004/05
sero-behavioural survey, as in Table 19 below.
Table 19: Adult HIV Prevalence, 2004/05 (%)

Age cohort M F
15-19 0.03 2.6
20-24 2.4 6.3
25-29 5.9 8.7
30-34 8.1 12.1
35-39 9.2 9.9
40-44 9.3 8.4
45-49 6.9 8.2
50-54 6.9 5.4
55-59 5.1 7.4
Source: Uganda HIV/AIDS Sero-Behavioural Survey, 2004/05

In the absence of the data about the effect of HIV/AIDS on fertility, the default estimates were used
(see Table 20). The model assumes that fertility is not affected by the provision of ART. In practice
this may not be the case, as ART could offset some of the negative impacts on fertility that are
associated with HIV infection. A study addressing this issue was conducted in Mbarara in 2005-06,
which tracked the fertility desire and history amongst 500 HIV+ women18. The study found that ART
use was associated with increased probability of fertility desire but decreased probability of

18
Emenyonu N, Maier M, Andia-Biraro I, Kaida A, Hogg R, and Bangsberg D: ART, Fertility Desire and Recent
Fertility History of HIV+ Ugandan Women

61

pregnancy and live birth. Hence there was no evidence that ART use increased actual (as opposed to
desired) fertility.
Table 20: Ratio of Fertility of HIV-Infected Women to the Total Fertility of Uninfected Women

Age Ratio
15-19 1.5
20-24 0.7
25-29 0.7
30-34 0.7
35-39 0.7
40-44 0.7
45-49 0.7
Source: Spectrum Model

The other default values from Spectrum used in the model are shown in Table 21, showing the
cumulative number of people dying of HIV/AIDS taking a slow pattern for both adult and children.
The average (median) time from infection to death without ART is 11 years for both male and female
adults, and approximately six years for children. It should be noted that this does not mean that all
HIV+ people will die after a fixed time; rather, there is a probability distribution of deaths, whereby
some will die quickly while others will survive for a long time even without treatment.
Table 21: Cumulative Percentage of People Dying from HIV/AIDS by Number of Years since
Infection, without ART

Year Male Female Children
1 0 0 25
2 0 0 34
3 1 1 39
4 4 4 43
5 7 7 47
6 11 11 49
7 17 17 51
8 24 24 53
9 32 32 55
10 41 41 57
11 50 50 58
12 58 58 59
13 65 65 60
14 71 71 62
15 76 76 63
16 82 82 64
17 85 85 65
18 87 87 66
19 87 87 67
20 87 87 68

Treatment Scenarios
Two different treatment scenarios were modelled.
Low ART: assumes that ART provision remains at relatively low levels, and peaks at around 45% of
the adult population with advanced HIV infection receiving ART. It also assumes that 90% of those
on ART at any time survive until the next year.

62

High ART: assumes that ART provision continues to grow steadily from current levels until 90% of the
clinically-eligible population is reached, and that 95% of those on ART at any time survive until the
next year.
A pilot study of treatment adherence and drug resistance conducted in Uganda between 1998 and
2000 gave encouraging results19. It found that as long as an effective drug supply chain could be
maintained, people in therapy reported good adherence, and viral and immune responses were
similar to those seen in North America and Europe.
The projections generated here assume that ART take-up does not vary between men and women
(i.e. that equal proportions of eligible HIV+ men and women take up treatment, and that adherence
is also independent of gender). While this may not be the case in practice, there is no reliable data
on different take-up and adherence rates by gender, and the Spectrum model does not disaggregate
ART treatment by gender.
The percentage of those with advanced HIV infection receiving ART under the two scenarios is
shown in Figure 29 below.
Figure 29: ART Rollout

Source: Own projections

4. Population Estimates
The estimates of the total population in the four scenarios are shown in Figure 30 below.
The results for all four scenarios are shown in this exercise (No AIDS, AIDS with No ART, Low ART and
High ART).
The results show that there was little difference in the projected population in the different
scenarios through the 1980s (which may indicate that the model projections are somewhat delayed

19
Weidle P Assessment of a pilot anti-retroviral drug therapy programme in Uganda: patients’ response,
survival and drug resistance.

63

in incorporating the initial impacts of HIV/AIDS, as AIDS-related deaths in the model during the
1980s are few).
Figure 30: Total Population

Source: Model projections

The “No HIV/AIDS” scenario puts Uganda’s total population in the year 2025 at 64.5 million. The age-
and gender-specific projections are as shown in Table 5 below. Putting HIV/AIDS into the projections,
without ART, a lower population of 58.8 million is estimated for 2025.
There are reasons to believe that Spectrum may be over-estimating the population. Under the
“With-AIDS” scenario, the projections show a total population of 26.3 million for 2002. This
compares with the observed census result for 2002, of 24.2 million people. It is not clear why this is
so, but suggests (if the census results are correct) that Spectrum is over-estimating the population by
some 10%. However, the main purpose of the present exercise is to compare projections under the
different scenarios (With/Without AIDS, etc.), and there is no reason to believe that comparative
projections are misaligned. These suggest that by 2002, HIV/AIDS had caused the Ugandan
population to be some 6% smaller than it would have been without HIV/AIDS, while by 2025 the
difference would be 9% (see Figure 31).

64

Figure 31: Population Deficit due to HIV/AIDS


Population Growth: The projections also show the impact of HIV/AIDS on population growth (see
Figure 32). These show that the main impact of HIV/AIDS was felt during the early 1990s, when
prevalence rates were relatively high. As prevalence rates fell during the late 1990s, population
growth is estimated to have risen, such that in the High ART scenario, it is almost up to the growth
rate projected for the “Without AIDS” scenario.
The projections show that even in the absence of HIV/AIDS, population growth declined during the
late 1990s. This is due to a combination of factors, including disruptive internal population
movements and out-migration (refer to Figure 26), as well as the expected natural decline in fertility
from very high rates.

65

Figure 32: Population Growth Rates


Table 22: Population Projections by Age and Gender for 2025 (million)

No HIV/AIDS HIV/AIDS (No ART)
Age group Total Male Female Total Male Female
0-4 12.22 6.16 6.06 11.24 5.66 5.58
5-9 9.94 5.00 4.94 9.20 4.63 4.57
10-14 8.49 4.26 4.22 7.89 3.96 3.93
15-19 7.09 3.55 3.54 6.59 3.30 3.29
20-24 5.68 2.84 2.84 5.27 2.64 2.63
25-29 4.66 2.32 2.33 4.33 2.17 2.16
30-34 3.84 1.91 1.93 3.60 1.81 1.78
35-39 3.09 1.53 1.55 2.86 1.44 1.41
40-44 2.43 1.20 1.23 2.19 1.10 1.08
45-49 1.95 0.96 0.99 1.68 0.85 0.83
50-54 1.55 0.76 0.79 1.27 0.64 0.62
55-59 1.20 0.58 0.62 0.91 0.45 0.46
60-64 0.87 0.42 0.45 0.64 0.30 0.34
65-69 0.62 0.29 0.33 0.46 0.20 0.25
70-74 0.42 0.19 0.22 0.32 0.14 0.18
75-79 0.25 0.11 0.14 0.21 0.09 0.12
80+ 0.21 0.09 0.12 0.19 0.08 0.11
Total 64.50 32.20 32.30 58.82 29.48 29.35

66

The projections also show the impact of providing ART on the total population. As Figure 31 above
shows, the provision of ART – even in the High ART scenario – only closes part of the population gap
between the “No AIDS” and “With AIDS” scenarios. In the Low ART scenario, the population in 2025
is only 0.1% higher than in the “No ART” scenario, while in the High ART scenario the population is
0.8% higher than in the “No ART” scenario.
The reason for the apparently small impact of ART provision is that a large proportion of the impact
on the total population was felt during the late 1980s and the early 1990s, where high HIV-
prevalence and death rates had a permanent effect, making the population smaller. The projections
also show that unless ART is widely provided, it pays little demographic dividend.
Table 23: Population Projections by Age and Gender for 2025 (million) – “With ART” Scenarios

Low ART High ART
Age group Total Male Female Total Male Female
0-4 11.25 5.67 5.58 11.31 5.7 5.61
5-9 9.21 4.63 4.58 9.25 4.66 4.6
10-14 7.9 3.97 3.93 7.94 3.99 3.95
15-19 6.59 3.3 3.29 6.61 3.31 3.3
20-24 5.27 2.64 2.63 5.27 2.64 2.64
25-29 4.34 2.17 2.16 4.35 2.17 2.17
30-34 3.6 1.81 1.79 3.63 1.82 1.81
35-39 2.87 1.45 1.42 2.91 1.47 1.45
40-44 2.2 1.11 1.09 2.25 1.14 1.12
45-49 1.69 0.86 0.83 1.74 0.88 0.86
50-54 1.27 0.64 0.63 1.29 0.66 0.64
55-59 0.91 0.45 0.46 0.92 0.46 0.47
60-64 0.64 0.3 0.34 0.64 0.3 0.34
65-69 0.46 0.2 0.25 0.46 0.21 0.25
70-74 0.32 0.14 0.18 0.32 0.14 0.18
75-79 0.21 0.09 0.12 0.21 0.09 0.12
80+ 0.19 0.08 0.11 0.19 0.08 0.11
Total 58.90 29.51 29.39 59.31 29.70 29.61

5. Further Analysis of the Impact of HIV/AIDS and the Provision of
ART
HIV Population: As can be seen in Figure 33, the number of people infected with HIV is estimated to
have peaked at about 1.4 million in 1996, before falling slowly. Without ART, the number of HIV+
people would continue to decline through to about 2012, following which time it would start to rise.
This reflects a number of factors. First, population growth – even with a constant prevalence rate, if
the population is growing then the number of those infected with HIV will rise. Second, there are
indications that the prevalence rate has been rising slightly (see Figure 28 and Figure 34), which
reinforces the upward trend in numbers infected.
With ART, the increase in the numbers of HIV+ people is even more dramatic, especially in the High
ART scenario. The rollout of ART increases the number of HIV+ people, as those who would have
earlier died are now living longer. The striking impact of this is shown in Figure 33. By 2025 there are
projected to be 2.2 million HIV+ people under the High ART scenario, but only 1.8 million in the
absence of ART.

67

Figure 33: Number of People Infected with HIV/AIDS


HIV Prevalence: Figure 34 shows the impact of ART provision on HIV prevalence. Although it is
projected that there is an underlying trend of declining prevalence, the impact of ART provision on
keeping people alive raises the overall prevalence rate. In the absence of ART provision, the adult
HIV prevalence rate is projected to fall to 5.1% in 2025. With ART, however, the prevalence rate is
projected to be higher, at 6.2%.
Figure 34: Adult HIV Prevalence


68

Number Receiving ART: The number of people receiving ART continues to rise in both the Low and
High ART scenarios, although much more dramatically in the latter (see Figure 35). The number
receiving ART in the High scenario is close to, but somewhat below, the projections contained in the
NSP. This may indicate that the model is under-projecting the number of HIV+ people, or that the
NSP envisages earlier treatment of HIV+ people with ART than the protocols embedded in the
Spectrum model. It is unlikely to reflect a faster rollout of ART in the NSP, as the High scenario
envisages a very rapid rollout of ART.
Figure 35: Number of Adults Receiving ART


AIDS-related Deaths and Life Expectancy: Figure 36 shows that the number of deaths as a result of
HIV/AIDS has been falling since the late 1990s, reflecting the earlier decline in HIV prevalence. Going
forward, the number of projected AIDS-related deaths is highly dependent upon the rollout of ART.
Under the High ART scenario, the number of AIDS-related deaths is projected to keep falling steadily
until around 2011, when ART provision levels out at 90% of the relevant eligible population. During
this period, the rapid rollout of ART dramatically cuts the number of AIDS-related deaths. After 2011,
the number of AIDS-related deaths starts rising again. It should be noted that the provision of ART
delays AIDS-related deaths but does not prevent them, due to various factors associated with ART,
including patients’ adherence lapses and the emergence of drug resistance. In the medium term,
however, it is clear that ART leads to a significantly reduced death rate and hence improved life
expectancy. Without ART (or in theLow ART scenario), the number of deaths is projected to decline
much more slowly, reflecting only the earlier decline in prevalence. Eventually, however, the number
starts rising, following the increase in the number of HIV+ people in the population.

69

Figure 36: AIDS-related Deaths


The overall impact of HIV/AIDS on life expectancy is shown in Figure 37. This shows that by the late
1990s, life expectancy had fallen to an estimated 44 years, compared to 56 years in the period
without HIV/AIDS. However, going forward, the gap declines, reflecting the decline in the HIV
prevalence rate and, in the High ART scenario, the availability of treatment that prolongs survival
times for HIV+ individuals. By 2025, life expectancy is projected to be 60 years in the “With-AIDS”
scenarios, compared to an estimated 64 years “Without AIDS”.

70

Figure 37: Life Expectancy


AIDS-related Deaths and the Labour Force: In populations without HIV/AIDS, death rates tend to be
higher for the elderly and very young children than for the general population as a whole. This
largely reflects the impact of disease relative to levels of health and bodily resistance to infection.
However, HIV/AIDS changes the pattern of deaths, and the age pattern of death rates are quite
different in a population that has high HIV prevalence than one without. HIV/AIDS tends to raise the
number of deaths amongst young and middle-aged adults, i.e. those who are economically most
productive and who are more likely to be skilled and employed. Hence investments made by the
government and other agencies to sustain the lives of the infected persons, will to some degree, be
balanced by the economic contributions of the same people. As shown in Figure 38, provision of
ART will have a significant effect on reducing AIDS-related deaths among the economically
productive age group of 15-59 years.

71

Figure 38: Age-specific AIDS-related Deaths in 2015


AIDS-related deaths are mainly concentrated in people of productive age, implying that Uganda has
been losing a large number of the country’s potential labour force (15-59 year old) since the late
1980s20. A sharp increase in these deaths was recorded in the 1990s, when the HIV prevalence was
at its peak. Thereafter, the death rate started declining as HIV prevalence declined.
In view of the potential economic impact of losing the most productive age group of the population,
the Spectrum model has been used to make projections of the likely loss of labour force that the
country may face in the period up to 2025. Under the different scenarios modelled, in the absence of
ART, the country will have lost 8.5% of the labour force (compared to the No-AIDS scenario) by 2025.
However, this can be reduced slightly (8.4%) in the low case scenario of ART. A bigger impact is
visible (7.7%) in the High ART scenario.

20
Although the Spectrum model has looked at the 15-59 population, this is driven by the model and
international conventions regarding the definition of the labour force. It does not mean that people do not
engage in economic activities before they are 15 years or when they are 60 years and above.

72

Figure 39: Projected Loss of Labour Force (15-59 yrs)


Orphans: The number of children orphaned to HIV/AIDS (both double and single orphans21) started
rising in the late 1980’s and steadily increased through the early 1990s. However, the rate of
increase tended to decrease in late 1990s, which could be attributed to decline in the HIV prevalence
rate at the time. The number of orphans has recently peaked at just over one million. Without the
intervention of ART, the total number of HIV/AIDS-orphaned children is estimated to be 830,000 in
2025. With ART, the number of orphans is expected to be lower. The Low ART scenario only leads to
small difference in the number of orphans, which would be 800,000 in 2025, whereas in the High
ART scenario the number would be much lower, at 620,000 (see Figure 40). For double orphans, the
numbers are much lower, peaking at around 400,000 and declining to under 200,000 by 2025.

21
Double orphans have lost both parents to HIV/AIDS while single orphans have lost either their mother or
father.

73

Figure 40: Orphan Projections (Double & Single)


6. Conclusions and Implications
The population projections detailed here do have some limitations. In particular, there is little
gender disaggregation related to the provision and uptake of ART, even though in practice the
behaviour of men and women may be quite different. This issue can be dealt with in future once
more disaggregated data are available and the Spectrum model can generate projections on this
basis. A further limitation is that the impact of ART on fertility is not accommodated, but again this
can be reflected in future once reliable data are available based on empirical research.
The main findings of this demographic study of HIV/AIDS in Uganda are as follows:
• The main demographic impact of HIV/AIDS has already occurred, i.e. during the 1980s and
early 1990s when HIV prevalence was very high, and there were large numbers of AIDS-
related deaths. As a result, the differences between the various scenarios going forward (e.g.
in terms of population deficit) are not very large. Going forward, however, there are
differences between the High ART and No/Low ART scenarios, in terms of the size of the
population, the numbers of HIV+ individuals, and the numbers of AIDS-related deaths.
However, the differences between the Low ART and No ART scenarios are minimal, and in
some cases indistinguishable, which indicate that a Low ART approach will yield few
demographic benefits (or social and economic benefits).
• In the High ART scenario, the number of AIDS-related deaths is cut over a fifteen year period
from around 2003 to 2018. Towards the end of the projection period, however, the number
of AIDS-related deaths is similar in both the High ART and No ART scenarios, essentially as a
“catch-up” process takes place. It is important to realise that ART does not keep all HIV+

74

individuals alive indefinitely, and that problems of insufficient adherence to treatment
regimens and emerging drug resistance will mean that some of those taking ART will
eventually die of AIDS-related illnesses. Importantly, however, they will have enjoyed many
extra years of fulfilling healthy life in the meantime.
• The widespread provision of ART (as in the High ART scenario) is a long-term, open-ended
commitment, which continues to grow during the period of the projections. Hence it is
important that efforts to prevent the spread of HIV/AIDS and reduce new infections
(incidence) are pursued, as this is the only long-term solution to the epidemic. It is especially
important that the heightened focus on treatment does not detract from the long-term
need for effective prevention.
• Life expectancy is higher in the High ART scenario, but does not recover to the levels that
would have been experienced without HIV/AIDS.
• The number of orphans (who have lost one or both parents to AIDS) peaked at just over one
million, or around 3% of the population. Providing ART cuts the number of orphans by
around one-quarter by 2025.
• By keeping HIV+ individuals alive for longer, the High ART scenario will lead to a higher HIV
prevalence rate, reinforcing the point that trends in HIV prevalence are not a good indicator
of the success or otherwise of HIV-prevention efforts in an environment of widespread ART
provision, and the focus has to be on incidence (not prevalence) rates.

75

Annex: Description of the Spectrum and
AIDS Impact Models 22
Spectrum Policy Modelling System
Spectrum is a computer programme for making population projections, but which is particularly
suited for modelling the impact of HIV/AIDS and the impact of policy interventions. It is based on the
analysis of existing information to determine the future consequences of various development
programmes and policies.
The Spectrum Policy Modelling System comprises an integrated package containing the following
components:
Demography (DemProj): A programme to make population projections based on (1) the current
population, and (2) fertility, mortality, and migration rates for a country or region.
AIDS Impact Model (AIM): AIM projects the consequences of the HIV/AIDS epidemic, including the
number of people living with HIV/AIDS, new infections, and AIDS-related deaths by age and sex as
well as new cases of tuberculosis and AIDS orphans.
BenCost: Financial benefits and costs of family planning programmes.
Allocate: Impact of resource allocation to different components of a reproductive health action plan.
Condom Requirements: A programme to forecast national condom requirements for both family
planning and HIV/AIDS prevention.
FamPlan: Projects family planning requirements needed to reach national goals for addressing
unmet need or achieving desired fertility.
NewGen: Reproductive health for adolescents.
PMTCT: Prevention of mother-to-child transmission.
RAPID: Projects the social and economic consequences of high fertility and rapid population growth
for such sectors as labour, education, health, urbanisation, and agriculture.
Safe Motherhood: Represents the relationships between a national maternal health programme and
the resulting maternal mortality ratio (MMR) and the number of maternal deaths.
This study makes use of DemProj and AIM. DemProj is at the heart of the Spectrum suite of models
as it is used to create the population projections that support many of the calculations in the other
components, such as FamPlan, Benefit-Cost, AIM, and RAPID.
The Spectrum Policy Models are designed to answer a number of “what if” questions. The “what if”
refers to factors that can be changed or influenced by public policy.

22
Based on the relevant software manuals, Stover & Kirmeyer (2005 and Stover (2005)

76

DemProj
The demographic model in Spectrum, known as DemProj, is a computer programme for making
population projections for countries or regions. The programme requires information on the number
of people by age and sex in the base year, as well as current year data and future assumptions about
the TFR, the age distribution of fertility, life expectancy at birth by sex, the most appropriate model
life table, and the magnitude and pattern of international migration. This information is used to
project the size of the future population by age and sex for as many as 150 years into the future. If
desired, and if suitable source data are available, the projection can also estimate the size of the
urban and rural populations.
Linking DemProj with other modules in Spectrum makes it possible to examine the demographic
impact of AIDS, the family planning service requirements to achieve demographic and health goals
(FamPlan), the costs and benefits of family planning programmes and the socioeconomic impacts of
high fertility and rapid population growth. DemProj was first produced in 1980. Since then, it has
been used by a large number of planners and researchers around the world. It has been updated
from time to time in response to comments and suggestions from users.

Data Inputs Required by DemProj
A. Base Year Population
The starting point for projections is the number of people in the population by age and sex in the
base year. For both males and females, the population is divided into five-year age groups from 0-4
to 75-79. There is also a final age group for those people aged 80 and older.
Base year population figures are available from a number of sources. Usually, the best source will be
a national census. There are other sources of population data if recent census reports are not
available. The Population Division of the United Nations publishes a considerable amount of
population data. The most useful sources for population projections are the Demographic Yearbook,
which contains the most recent census data for most countries; and the World Population Prospects,
published every two years and containing population estimates and projections for most countries
of the world. World Population Prospects contains estimates of base year populations as well as
assumptions about future levels of fertility, mortality and migration, including estimates and
projections of population by five-year age groups that have been adjusted for misreporting. These
data may be used when reliable census data are not available.
DemProj contains a module called EasyProj which use data from World Population Prospects to
produce national population forecasts from base year data.
When DemProj is to be used to provide inputs to the AIM, then the base year must be sufficiently far
back in history that HIV/AIDS had no impact on the population.
B. Fertility
A population projection requires information about the level of fertility (obtained through the TFR)
and about its shape (obtained through the age distribution).
1. The Total Fertility Rate
Base Year Estimates: The TFR is the number of live births a woman would have if she survived to age
50 and had children according to the prevailing pattern of childbearing at each age group. Estimates
of the TFR are available from a number of sources. The best sources will be national fertility surveys.

77

Future Assumptions: An assumption about the future TFR is required for most population
projections. Again, the best source is usually national projections or population policy goals.
Alternatively, projections may be obtained from international sources such as the United Nations
Development Programme (UNDP) as reported in the World Population Prospects.
2. The Age Distribution of Fertility
In addition to the TFR, the age distribution of fertility is also required to make a population
projection. In DemProj, this information is entered as the percentage of lifetime fertility that occurs
in the five-year age groups 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, and 45-49.
C. Mortality
Mortality is described in DemProj through two assumptions: life expectancy at birth by sex, and a
model life table of age-specific mortality rates.
1. Life Expectancy at Birth
Base Year Estimates: Life expectancy at birth is the average number of years that a cohort of people
would live, subject to the prevailing age-specific mortality rates. It is a useful measure that
summarises in one indicator the effect of age-specific mortality patterns. Life expectancy can be
calculated from vital statistics on deaths if reporting is complete. In the developing world, death
registration is not usually complete enough to be used for this purpose. Estimates of life expectancy
are usually derived instead from large-scale surveys or censuses. The best source of information on
life expectancy will usually be national reports prepared by analysing these surveys. If national
estimates are not available, life expectancy estimates may be obtained from a variety of other
sources, including the United Nation’s World Population Prospects or the Demographic Yearbook,
the USA Census Bureau’s World Population Profile, the World Population Data Sheet of the
Population Reference Bureau, or the World Bank’s World Development Indicators.
Future Assumptions: An assumption about future levels of life expectancy at birth is required for all
population projections. There are several options for setting the life expectancy assumption. These
include national projections, national population goals, United Nations and USA Census Bureau
projections, or recent trends and international experience, or the United Nations model schedule.
2. Life Expectancy and AIDS
In a number of countries, the AIDS epidemic has had a significant impact on mortality. It affects both
life expectancy and the age and sex pattern of mortality. This health concern raises two problems for
population projections. First, in countries with high HIV prevalence, the future course of the AIDS
epidemic will be the single largest determinant of future life expectancy. Second, the age pattern of
mortality will depart significantly from the patterns described in the model life tables discussed
below.
Therefore, in countries with adult HIV prevalence greater than a few percent, it is best to consider
the effects of AIDS explicitly in the population projection. These effects cannot be incorporated
simply by changing the life expectancy assumption since the age pattern of mortality is also affected
(AIDS-related deaths are concentrated in the age groups of 15-49). The recommended approach is to
first develop a population projection that ignores the effect of AIDS, then to make assumptions
about the future level of adult HIV prevalence and let the computer programme calculate the effects
of AIDS on the population projection. Such projections can be prepared using DemProj and AIM, the
AIDS component of Spectrum.

78

3. Age-Specific Mortality
Model Mortality Tables: The mortality input to DemProj, life expectancy at birth, indicates overall
mortality in a population. But Demproj also needs the pattern of mortality in order to produce
mortality rates by age group. Specifically, the rates required by DemProj are survival ratios, which
will survive one age group into the next five-year group.
The majority of countries to which DemProj has been applied have had no complete, empirical life
tables - and life tables are what yields survival ratios, or sx. Even if there were such tables, generally
little is known about how the pattern of mortality would evolve, given projected changes in
mortality levels. So DemProj employs model life tables: the Coale-Demeny and the United Nations
tables for developing countries (United Nations, 1982). Although these two sets differ in the
algorithm they use to generate the mortality schedules, and the empirical data sets from which they
were drawn, they are similar in that they contain regional families that are distinguished by
underlying causes of death. The applicable model life tables for a particular country can either be
derived from typical regional tables (if a country is similar to others in the region) or by comparing
the range of tables with actual data on key indicators such as life expectancy and crude death rates.
D. Migration
Migration refers to the number of migrants moving into or out of the area for which the population
projection is being prepared. If the projection is for a country, then it is international migration.
Migration is specified through two inputs. The first is the net number of migrants, by sex and year. If
the net flow is outward, then net migration should be a negative number. If the net flow is inward,
then the net migration should be a positive number. In most cases, information on migration will
come from local sources, usually studies based on a national census. The United Nations report
World Population Prospects does contain estimates and projections of total net migration, but they
are not disaggregated by sex.

Projection Outputs
DemProj will calculate and display the population size by year. Projections can be examined in terms
of total population or population by age, sex, and region. In addition, a number of demographic
indicators can also be displayed. A complete list of indicators available is given below.
Total population size
Population aged 0-4
Population aged 5-14
Population aged 15-64
Population aged 65+
Total net international migration
Annual Growth Rate (GR): The rate at which the population is increasing or decreasing in a given
year due to natural increase and net migration, expressed as a percentage of the base population.
Births: The total number of annual births.
Child-Woman Ratio: The number of children under the age of five per woman of childbearing age
(15-49).
Crude Birth Rate (CBR): The number of live births per 1,000 population in a given year.

79

Crude Death Rate (CDR): The number of deaths per 1,000 population in a given year.
Deaths: The total number of annual deaths.
Defined Age Group: The size of the population in a user defined age group.
Dependency Ratio: The ratio of the economically dependent part of the population (those aged 0-14
and 65 and over) to the productive part (those aged 15-64).
Doubling Time: The number of years it would take for the population to double its current size at the
current annual rate of growth.
Gross Reproduction Rate (GRR): The average number of daughters that would be born to a woman
(or a group of women) during her lifetime if she passed through all her childbearing years
conforming to the age-specific fertility rates of a given year. This is similar to the TFR except that it
counts only daughters.
Infant Mortality Rate (IMR): The number of deaths of infants under one year of age per 1,000 live
births.
Life Expectancy [e(0)]: The average number of years a person can expect to live based on the age-
specific death rates for a given year. This is the calculated life expectancy at birth. If AIM is not being
used, then this number will be the same as the input life expectancy. However, if AIM is being used,
then the calculated life expectancy will include the impact of AIDS-related deaths and will be
different from the input life expectancy.
Mean Age of Childbearing: The average age of mothers at the time of birth.
Median Age: The age that divides a population into two numerically equal groups.
Net Reproduction Rate (NRR): The average number of daughters that would be born to a woman (or
a group of women) during her lifetime if she passed through all her childbearing years conforming to
the age-specific fertility rates and age-specific mortality rates of a given year. This is similar to the
GRR except that it includes the effect of mortality that would cause some women to die before
completing their childbearing years.
Rate of Natural Increase (RNI): The rate at which the population is increasing or decreasing in a
given year due to the surplus or deficit of births over deaths, expressed as a percentage of the base
population.
Sex Ratio: The number of males per 100 females in a population.
Total Fertility Rate (TFR):. The average number of children that would be born alive to a woman (or
a group of women) during her lifetime if she were to pass through all her childbearing years
conforming to the age specific fertility rates of a given year.
Under-five Mortality Rate (U5MR): The number of deaths to children under the age of five per
1,000 live births.

Methodology
Details of the methodology used to prepare the projections can be found in the Spectrum manual
(Stover & Kirmeyer, 2005).

80

AIDS Impact Module (AIM)
Projection Inputs
AIM requires data describing the characteristics of the HIV/AIDS epidemic and the response to it.
Some of these data (e.g., adult HIV prevalence) must be specific for the area being studied, whereas
others (e.g., the MTCT rate) can be based either on local data or on international averages when
local data are unavailable. This section describes the inputs required, possible sources and
recommendations for default values when local data are not available.
A. Demographic Projection
As noted above, AIM requires that a demographic projection first be prepared using DemProj. Two
key points are relevant when preparing a DemProj projection for use with AIM:
1. For accuracy, the first year of the projection should be before the starting year of the HIV/AIDS
epidemic. It is possible to start the projection in a year after the beginning of the AIDS epidemic, but
this type of projection will be less accurate.
2. The life expectancy assumption entered into DemProj should be the life expectancy in the
absence of AIDS. AIM will calculate the number of AIDS-related deaths and determine a new life
expectancy that incorporates the impact of AIDS. It is necessary to use this two-step process because
model life tables (for specifying the age distribution of mortality) do not contain patterns of
mortality that reflect the excess deaths caused by AIDS.
B. Adult HIV Prevalence
Base Year Estimates: Adult HIV prevalence is the percentage of adults aged 15 to 49 who are
infected with HIV. Thus, this estimate of prevalence refers to the entire adult population aged 15 to
49, not just a specific risk group. The best source of prevalence data is from national Sero-Prevalence
Surveys. Sentinel surveillance surveys based on pregnant women can also be used, but prevalence
rates obtained from such surveys may not be representative of all adults. UNAIDS also prepares
estimates of national HIV prevalence for most countries, based on careful consideration of the
available surveillance data, by risk group; recent trends in HIV infection; and national population
estimates. The latest estimates are available from the UNAIDS website at http://www.unaids.org.
Future Projections: An AIM projection requires an estimate of future levels of HIV prevalence.
Usually AIM is used to illustrate the future consequences of an epidemic. Therefore, it is not
necessary to try to predict future prevalence. Rather, AIM can be used with plausible projections of
future prevalence to show what would happen if prevalence followed the indicated path. In this case
it is only necessary to have a plausible projection.
When AIM is used to stimulate policy dialogue, it is often helpful to use a conservative projection of
future prevalence. This approach will avoid charges that the presentation is using the worst possible
assumptions to make the case for AIDS interventions stronger and will allow the discussion to focus
on other, more important issues.
C. Progression from HIV Infection to AIDS Death
The progression period describes the amount of time that elapses from the time a person becomes
infected with HIV until he or she dies from AIDS. AIM uses the cumulative distribution of the
progression period. This distribution is defined as the cumulative proportion of people infected with
HIV who will die from AIDS, by the number of years since infection. For example, it might be that for

81

all people infected in a certain year, 1% will die within one year, 3% will die within two years, 7%
within three years, etc. The incubation period can be specified for up to 20 years. The cumulative
percentage dying from AIDS by year 20 will be the percentage that ever dies from AIDS. Thus, if this
value is equal to 95%, it implies that 5% of people infected with HIV will never die from AIDS. AIM
uses separate progression periods for adult men, adult women and children. AIM allows for “fast”
and “slow” patterns of incubation.
D. Age and Sex Distribution of Infections
To calculate HIV incidence from the prevalence input, AIM needs to have some information on the
distribution of infection by age and sex. This information is provided through two editors, one for
the ratio of prevalence at each age group to prevalence in the 25-29 age group, and one for the ratio
of female to male prevalence.
AIM has two default patterns, one for generalised epidemics and one for low level and concentrated
epidemics. Default patterns of the distribution of HIV infection by age by type of epidemic have been
developed from population-based surveys and reported AIDS cases.
Where population prevalence data are available for a particular country, the observed pattern can
be substituted for the default pattern in the survey year and the entire pattern over time will be
adjusted to match.
E. Mother-to-Child Transmission
The MTCT rate is the percentage of babies born to HIV-infected mothers who will be infected
themselves. Studies have found that this percentage ranges from about 13% to 32% in industrialised
countries and 25% to 48% in developing countries. The higher rates have generally been found in
studies in Africa, where a significant amount of transmission through breastfeeding may take place,
and the lower figures have been found in Western Europe. AIM uses a default value that depends on
breastfeeding practices. If country-specific information is available, it can be used instead of the
default values.
The probability of transmission may be changed if a country implements programmes to prevent
MTCT of HIV. The effect of such programmes can be included by indicating the type of treatment
used and the infant feeding options promoted. The type of programme can include treatment with
Nevirapine, AZT or some other treatment, as well as infant feeding options (formula feeding,
exclusive breastfeeding or mixed feeding).
F. Total Fertility Rate Reduction
It is not clear how the TFR might be affected by an HIV/AIDS epidemic. Some women who find that
they are infected with HIV may want to have as many children as possible while they can, in order to
leave descendants behind. Others may decide to stop childbearing upon learning that they are HIV+
in order to avoid leaving motherless children behind. Since the majority of people do not know if
they are infected or not, knowledge of HIV infection is not likely to have a large effect on the desired
fertility rate.
AIDS could lead to higher age at first intercourse as the dangers of unprotected sex become known.
This trend would lead to lower fertility rates. Studies of the determinants of fertility found no clear
evidence but concluded that the most likely result is that an HIV epidemic will slightly reduce
fertility. Two studies in Uganda found that HIV-infected women had lower fertility rates than HIV-
negative women. Since most women did not know their sero-status, the reduced fertility rates were

82

most likely due to biological rather than behavioural factors. This finding suggests that fertility might
be 20% to 50% lower among HIV-infected women.
The default value in AIM is that fertility among 15-19 year old women is 50% higher among HIV+
women than HIV-negative women and that fertility among women 20-49 is 20% lower among HIV+
women than HIV-negative women.
G. Anti-retroviral Therapy
Anti-retroviral therapy can extend life and improve the quality of life for many people infected with
HIV. ART has restored health to many people and continues to do so after many years. But ART does
not help everyone. Some people have a good reaction initially but over time the virus becomes
resistant to the drugs and the benefits diminish. Others experience such severe side effects that they
cannot continue to take the drugs.
AIM can calculate the effects of ART based on an assumption about the proportion of those in need
receiving ART. ART is assumed to delay progression to death as long as it is effective. However, some
people will develop resistance to anti-retrovirals and others may have to stop treatment because of
severe side effects. As a result, only a proportion of those on ART in one year continue the next year.
When a person stops ART, s/he progresses to die of AIDS-related illnesses quickly.
Since people with HIV will survive longer if they are on ART, introducing anti-retrovirals will tend to
raise prevalence initially as new infections continue to occur and there are fewer deaths. In most
cases the prevalence input will be derived from surveillance data collected when anti-retrovirals
were not available. Thus, both the prevalence input, and the resulting incidence estimate, can be
considered to represent the situation without ART. In that case, and if incidence remains the same,
introducing ART will raise prevalence above the input projection. However, if ART is already being
supplied to significant portions of the population, the historical surveillance data and, thus, the
prevalence projection input will already include the effect of ART. In this case, the prevalence
estimate should not be changed by ART; instead, incidence should be adjusted downward to
compensate for the life-prolonging effects of ART.
AIM determines the effect of ART on prevalence by comparing the coverage of ART in 2005 with the
highest coverage level in the years after 2005. If coverage increases significantly, then prevalence
will be affected by the longer survival of those on ART. If coverage is already at or near its maximum
value in 2005, then prevalence after 2005 will not be affected.
H. Orphans
AIM will estimate the number of AIDS and non-AIDS orphans caused by adult deaths. An orphan is
defined as a child under the age of 18 who has lost at least one parent. These estimates are based
on the time history of fertility and the age at death. AIM will estimate maternal orphans (children
whose mother has died), paternal orphans (children whose father has died), and dual orphans
(children whose father and mother have both died). AIDS orphans are children who have lost at least
one parent to AIDS. To estimate double AIDS orphans, AIM needs to estimate the proportion of
couples with both parents infected with HIV. This estimation is based on a regression equation using
data from national population surveys in SSA. To make the estimate more precise, two additional
pieces of information are required: the percentage of women aged 15- 19 who have not married,
and the percentage of married women who are in monogamous unions. Both of these parameters
are available from national population surveys for most countries.

83

Projection Outputs
A complete list of the indicators calculated and displayed by AIM and their definitions is given below.
A. Total Population
Number Infected with HIV: The total number of people who are alive and infected with HIV.
HIV Age Distribution: The number of infected people, by age and sex.
Number of HIV+ Pregnant Women: The number of pregnant women who are infected with HIV.
Number of new HIV infections: The total number of new HIV infections each year.
Adult HIV Incidence: The percentage of uninfected adults who become infected in each year.
New Infections by Age: The number of new infections by age and sex and incidence by age and sex.
New AIDS Cases: The number of people progressing to AIDS each year.
AIDS Age Distribution: The number of people alive with AIDS, by age and sex.
AIDS-related Deaths: The annual number of deaths due to AIDS.
Cumulative AIDS-related Deaths: The cumulative number of AIDS-related deaths since the beginning
of the projection.
AIDS-related Deaths by Age: The number of AIDS-related deaths each year by age and sex.
HIV/AIDS Summary: A table with a selection of indicators shown for a selection of years.
B. Adults (15-49 years old)
HIV Population: The total number of adults who are alive and infected with HIV.
Adult HIV Prevalence: The percentage of adults (population aged 15 to 49) who are infected with
HIV.
Number of New HIV Infections: The total number of new adult HIV infections each year.
Adult HIV Incidence: The percentage of uninfected adults who become infected in each year.
New AIDS Cases: The number of adults progressing to AIDS each year.
AIDS-related Deaths: The annual number of adult deaths due to AIDS.
Number Newly Needing ART: The number of adults progressing to the stage where they need ART.
This is estimated as those within two years of death from AIDS-related illnesses if they do not
receive ART.
Total Number Needing ART: The total number of people needing ART. This includes those newly
needing therapy and those who continue successfully on therapy from the previous year.
Number on ART: The number of people receiving ART.
Unmet need for ART: The number needing ART who are not receiving it.
Adult Population: The number of adults between the ages of 15 and 49.
Adults 15-49 Summary: A table showing indicators just for adults 15-49.
C. Children (0-14 years old)

84

HIV Population: The total number of children who are alive and infected with HIV.
Number of New HIV Infections. The total number of new child HIV infections each year.
New AIDS Cases: The number of children progressing to AIDS each year.
AIDS-related Deaths: The annual number of child deaths due to AIDS.
Children 0-14: The number of children between the ages of 0 and 14 years old.
Number of Children Needing ART: The number of children who have progressed to moderate-to-
severe disease and, therefore, need ART.
Number Receiving ART: The number of children receiving ART.
Child Summary: A table showing indicators just for children under the age of 15.
D. AIDS Impacts
Tuberculosis (TB) Cases: The annual number of new TB cases.
Young Adult (15-49) Deaths: The total number of annual deaths occurring to adults between the
ages of 15 and 49, inclusive.
E. Orphans
Maternal AIDS Orphans: Children under the age of 15 who have lost their mother to AIDS.
Paternal AIDS Orphans: Children under the age of 15 who have lost their father to AIDS.
Dual AIDS Orphans: Children under the age of 15 who have lost both parents to AIDS.
All AIDS Orphans: Children under the age of 15 who have lost one or both parents to AIDS.
Maternal non-AIDS Orphans: Children under the age of 15 who have lost their mother due to causes
other than AIDS.
Paternal non-AIDS Orphans: Children under the age of 15 who have lost their father due to causes
other than AIDS.
Dual non-AIDS Orphans: Children under the age of 15 who have lost both their parents due to
causes other than AIDS.
All non-AIDS Orphans: Children under the age of 15 who have lost one or both parents due to
causes other than AIDS.
Maternal Orphans: Children under the age of 15 who have lost their mothers due to other causes.
Paternal Orphans: Children under the age of 15 who have lost their father due to other causes.
Dual Orphans: Children under the age of 15 who have lost both their parents due to other causes.
Total Orphans: Children under the age of 15 who have lost one or both parents due to other causes.
Summary by Age: A table showing orphans by type and single age.
Summary Table: A table showing all orphans by type and year.

85

Chapter 6: An Empirical Analysis of
the Link between Aid
Flows, the Exchange Rate
and Inflation in Uganda
1. Introduction
Like many low-income countries in Africa, Uganda is heavily dependent on external donor support to
finance budgeted government spending. Donor funds have accounted for an average of 43% of total
public expenditures over the past decade, a large portion of which is spent on non-traded goods and
services. Over this period, Uganda has experienced a surge in aid inflows, reflecting the country’s
qualification for the Heavily Indebted Poor Countries (HIPC) debt relief initiative, and because of the
need to control the daunting scale of HIV/AIDS epidemic. Uganda was one of the ten top recipients
of total Official Development Assistance (ODA) flows to the health sector over the same period, and
over the last five years, Uganda received over US$ 400 million of aid to the Health sector under
donor project funding and US$ 496 million under the PAF budget. Uganda is also amongst the
world’s top ten recipients of aid earmarked for HIV/AIDS control. Between 2003/04 and 2006/07,
the national HIV/AIDS budget grew dramatically, from about US$ 40 million in 2003/04 to nearly US$
170 million in 2006/07, and is projected to rise even further. Almost all of this is provided by external
donors. The USA is the main donor, as Uganda is one of the focus countries of the PEPFAR. UNAIDS
and other UN Agencies which support a variety of HIV/AID activities in Uganda, as well as many
other donors as detailed in Chapter 4 of this report. Projections in the Uganda AIDS Commission
(UAC) National Strategic Plan for 2007/08 to 2011/12 indicate that total spending could double
further, to US$ 511 million in 2011/12, and this would be largely (85%) donor funded (UAC, 2007).
Such increased resource inflows should in principle be beneficial for Uganda, as in other low-income
developing countries. The additional resources can be used to assist in the treatment and prevention
of HIV/AIDS (as well as other diseases) and for addressing some of the social consequences of the
disease, as well as provide for the rapid replacement of human resources lost to the infection and
for the training of additional personnel to address the urgent issues of prevention, treatment and
mitigation.
However, the surge in aid inflows to Uganda, in particular towards HIV/AIDS programmes, is looked
on with some fear by many donors and policy makers due to its possible macroeconomic
consequences that may undermine growth. The concern about macroeconomic instability derives
from the fact that because aid flows are (i) large relative to the economy and (ii) often volatile and
sometimes politically influenced, they may have the potential to increase macroeconomic instability
(i.e. inflation, exchange rates as well as interest rates). Based on the presumption that a stable
macroeconomic environment is critical for growth, Uganda is committed to maintaining single digit
inflation levels, low budget deficits and BoP equilibrium. Among others, there is a fear that monetary
injections arising from additional aid-induced spending would lead to inflationary pressures and
volatility or appreciation of the REER unless sterilised. Moreover, rapid increases in aid (even when
it is sterilised) as in the case of Uganda, can lead to adjustment costs. First, the direct impact of large

86

aid inflows can affect the quality of aid management, coordination and public service delivery.
Second, there can be indirect impacts on key markets and sectors, including the labour market, the
capital goods sector and, money and foreign exchange markets. In the latter case, the volume of aid
flows to government may overwhelm the capacity of the domestic authorities to avoid short-run
volatility in the exchange rate and interest rates, both of which can be damaging to private sector
investment. This is particularly acute in countries where financial markets are thin. The
macroeconomic impact will also be larger to the extent that aid flows are spent on locally-produced
goods and services (rather than imports), and especially when spent on non-traded goods and
services.
The second area of concern is that receipts of foreign exchange can cause the nominal and real
exchange rate to appreciate, which may in turn adversely affect exports, economic growth,
diversification, employment and poverty.
These concerns also reflect the fact that various measures of Uganda’s exchange rate have shown a
tendency to appreciate during some periods, although in general the exchange rate has depreciated
in nominal and real terms over the past decade. In particular, the REER appreciated between
1992/93 qnd 1997/98, and again in 2001/02 and 2004/05, while the Nominal Effective Exchange
Rate (NEER) also appreciated in 1992/93 to 1994/95, 2001/02 and 2004/05 (see Figure 41). The
appreciation of the exchange rate has led to concerns over its possible effect on the prospective
competitiveness of Uganda’s exports (the Dutch disease)23, its possible long run effects on growth of
the economy24, and the reasons behind the appreciation, in particular the role played by aid inflows
to Uganda which increased rapidly in particular over the last decade to support Uganda’s poverty
reduction programmes.25
While these concerns have been under debate, attempts to empirically quantify them have been
inconclusive. This chapter contributes to the debates by offering some empirical evidence on two of
the major sets of macroeconomic linkages from increased aid flows. First, it considers the possible
impact of increased aid flows on the REER and inflation. Although the major concern for the broader
study is with the effects of aid devoted to HIV/AIDS programmes, we consider the effects of total aid
to the GoU, largely reflecting the nature of available data26. Using a Vector Autoregression (VAR)
model of Uganda, and monthly data for the period July 1994 to June 2007, we examine the
following: (i) The link between aid flows and inflation and (ii) The link between aid flows, inflation
and the REER. The results provide some conclusions regarding the possible impacts of aid on the
economy, and policy advice on whether increased aid towards welfare improvement is appropriate.

23
See for example, BOU/MFPED (2004).
24
The real appreciation reduces the competitiveness of the domestic traded goods sectors. Over the long
run, production in these sectors contract and resources shift to the production of non-tradables. This may
lead to a less diversified and more vulnerable economy that is increasingly dependent on external resource
flows.
25
External Budget support to Uganda rose from about 3 per cent of GDP in 1999 to about 8 % of GDP in 2001,
following the development of the Poverty Eradication Action Plan (PEAP) which led to the introduction of
the Poverty Action Fund (PAF) in 1998/99 (Atingi Ego, 2005).
26
While this may appear to be a potential problem with the analysis, it should not be. First, the available data
for such analysis only relates to total aid, and data on the resources specifically for HIV/AIDS are not
available over the time period and with the frequency (quarterly) that is required. Second, the
macroeconomic effects considered here (on inflation, the exchange rate, exports etc.) result from the
magnitude of the incoming financial flows rather than what they are earmarked for, so the analytical
results in this chapter do not depend on the nature of the aid, just its magnitude.

87

Second, the chapter quantifies the relationship between the exchange rate and exports, in order to
provide insights about the possibility of an aid induced Dutch disease. In particular, the study
attempts to answer the following question: “What is the nature and extent of the effects of changes
in REER on individual goods and services exports in Uganda?” The result could be used to gauge
whether there has been an aid-induced Dutch disease effects on Uganda. It also addresses whether
exchange rate volatility affects exports in Uganda, and draws out policy lessons that can be learnt
from the empirical findings.
To accomplish these objectives, we estimate a model of the determinants of Ugandan exports
supply over the period 1993-2006. We focus on six exports: three traditional exports (coffee, cotton
and tea) and three non-traditional exports (fish, maize and flowers). The merchandise exports
selected together accounted for 44% of total export receipts in 2005/06.
The rest of the chapter is structured as follows. Section 2 provides a brief literature survey on the
aid-macro debates, and on the nexus of ODA, exports and exchange rates. Section 3 provides a
descriptive analysis of the linkages between aid inflows, inflation and the REER, and of the trends in
selected determinants of Ugandan export supply. This is followed in Section 4 by a description of the
methodology and the data. The empirical findings are discussed in Section 5. Section 6 concludes
and draws policy implications.
Figure 41: Nominal and Real Effective Exchange Rates

Source: Bank of Uganda

2. Literature Survey
Aid Flows, Exchange Rate and Inflation
Financial inflows carry some benefits to the recipient countries. As noted by several writers (for
example, Serieux, 2007), the additional resources can be used to: (i) Provide badly needed resources
for the treatment and prevention of HIV/AIDS (as well as other diseases) and for addressing some of
the social consequences of the disease; (ii) Provide for the rapid replacement of human resources
lost to the infection and the training of additional personnel to address the urgent issues of

88

prevention, treatment and mitigation; (iii) Make progress towards the achievement of other MDGs;
and (iv) Provide relief from savings and foreign exchange constraints, thus allowing for more optimal
(and ultimately more growth-enhancing) decision making with respect to production and
investment-related choices.
However, the effectiveness of large inflows of ODA requires the ability of the recipient countries’
institutions and macroeconomic policies to stimulate private sector investment. Therefore, in
assessing the benefits with which ODA is associated, they should be evaluated against its possible
harmful side effects on the REER, prices and the adjustment costs. Various literatures point out the
aid benefits-aid costs nexus.
Nkusu (2004a), for instance, argues that, conscious of the possible adverse effects of the ODA
inflows on the REER, policy makers have responded by sterilising (sale of Treasury bills) excess
liquidity injections arising from government spending of the ODA. But, the high domestic interest
rates that sterilisation induces creates macroeconomic management problems by attracting further
capital inflows especially for countries that are integrated to the global financial markets, exerting
further pressures on the REER.
Indeed, in a situation where the government wishes to maintain the existing rate of growth of
money supply, foreign exchange reserves and domestic credit to the private sector, the monetary
expansion arising from additional aid spending would need to be sterilised by sales of foreign
exchange or Treasury bills and bonds to limit aid-spending induced growth of money supply. But, the
effect of aid spending also depends on how the government chooses to split its public expenditures
between imports, domestically produced tradable goods and non-tradable goods and services.
Sanjeev, Powel and Yang (2005) argue that the macroeconomic impact of scaling up aid depends on
how aid is spent, its composition and its assumed policy response. They argue that it is the
interaction of the government’s fiscal policy with monetary and exchange rate management that
matters. If aid resources are spent directly by government on imports or if aid is spent in-kind (for
example on drugs), there is no direct impact on the exchange rate, price level, or interest rates. But,
if the government immediately sells to the central bank the foreign exchange it receives, then it
must decide how much of the local currency counterpart to spend domestically, while the central
bank must decide how much of the aid-related foreign exchange to sell on the market. In general,
this would impact on the exchange rate, the price level and interest rates.
If aid resources are spent directly on domestically produced non-tradable goods and services, it will
increase domestic demand, thereby inducing rising prices of the non-tradable goods and spilling over
into generalised inflation, and an appreciation if not fully offset by productivity-enhancing supply
side effects which are associated with higher aid flows. The extent of the effect of increased
spending depends on the income elasticity of demand and price elasticity of supply in the domestic
economy.
Adam and Bevan (2002) and Nkusu (2004b) note that the more elastic the supply response, the
smaller the required contraction in private demand (smaller REER appreciation), while the more
income elastic the demand for non-tradable goods is, the larger the appreciation of the REER could
result from increased government spending. They argue that in the absence of spare capacity, the
scale of the induced appreciation will depend partly on the composition of government spending
and partly on the extent of substitutability in private demand. The existence of excess capacity in the
tradable sector will tend to increase the appreciation, while excess capacity in the non-tradable

89

sector will tend to reduce it. For further discussions see, Tsikata (1999), Hansen and Tarp (2000),
Collier and Dehn (2001) and Collier and Dollar (2002).
Another but less pronounced concern about the effect of rapid increases or high levels of aid flows
has been volatility of aid. Aid flows are volatile for several reasons (and in several ways). First, the
information content of aid commitments (the quantity generally used in recipient countries’
budgeting), with respect to actual disbursements, is poor. Second, even aid disbursements
themselves are more volatile than fiscal revenues. Third, aid is generally procyclical relative to
revenue, meaning that it tends to exacerbate the variability in revenue streams. An implication of
volatility of aid is that, in and of itself, a sharp unanticipated change in the amount of aid received by
a country that is credit-rationed in international capital markets (the case for most high-aid
recipients) is, effectively, an exogenous shock that imposes adjustment costs on the economy.
The marginal cost of aid (i.e. the macroeconomic distortions it imposes on the economy) is likely to
rise with the rate at which the aid flow is increased. Hence, too rapid an increase in aid can reduce
its effectiveness at the margin and lower the level of aid that can be absorbed before it starts to
have an overall adverse impact on the economy. Adjustment-cost concerns emerge in two areas.
First, through the direct impact of large aid inflows on the quality of aid management, coordination
and public service delivery (so-called ‘micro absorption’ constraints). Second, they reflect indirect
impacts via key markets and sectors:
• The labour market: Where the demand for labour in critical sectors can only be met through
higher labour costs and lower skill levels. These pressures may occur in the public sector, both at the
‘implementation’ end (e.g. in the need for doctors) and at the ‘coordination and management’ end
(e.g. in the Ministry of Health), but may also be felt elsewhere in the labour market if the public
sector drives up wages or ‘cherry picks’ skilled workers (DFID, 2004).
• The capital goods sector: Where increased demand for investment in domestic assets
(construction goods) caused by short-run exchange rate appreciation, raises their price and reduces
the marginal efficiency of investment.
• The money and foreign exchange markets: Where the volume of aid flows to government may
overwhelm the capacity of the domestic monetary authorities to avoid short-run volatility in the
exchange rate and interest rates; both of which are damaging to private sector investment. This risk
is particularly acute in countries where financial markets are thin.
As indicated by Pallage and Robe (2001), the welfare costs of the business cycles created by these
shocks are particularly high in low-income countries. Further, when the pro-cyclicality of aid is added
to its own cycle-inducing effects, that cost is further magnified. In cases where a cash budget is used
to manage public sector spending, the stop-start-stop effect induced by the volatility of aid further
compromises the effectiveness of public sector activity, with concomitant welfare and growth costs
(Bulir and Hamann, 2003). In short, high aid-receiving countries face the very real prospect of
greater volatility in fiscal outcomes and economic activity and reduced public sector effectiveness.
This chapter also examines the evidence of macroeconomic effects of aid in Uganda focusing on
inflation and REER. The focus is on aid that is channeled through the government, as a proxy of total
aid to Uganda because data on unofficial aid is not available as a continuous time series. However, it
is acknowledged that currently a large proportion of the aid coming in to support the fight against
HIV/AIDs, for example, aid under PEPFAR is not captured in the official aid statistics reported by
government although its macroeconomic impacts would be similar to that of the official aid,

90

provided that the aid is spent domestically. As Report Number 3 of this project has noted, only about
40% of the aid from unofficial sources towards HIV/AIDs issues is spent domestically. This could have
macroeconomic impacts (both positive and negative ones).

Real Exchange Rate and Exports
At face value, increased external development aid for Uganda, like many other developing countries,
would appear to be nothing but good news. For instance, several writers e.g. (Serieux, 2007), argue
that additional resource inflows can be used to: (i) Provide badly needed resources for the treatment
and prevention of HIV/AIDS as well as other diseases and for addressing some of the social
consequences of the disease, such as the care of AIDS orphans and the repair and support of the
challenged or compromised community institutions and household structures; and, (ii) Provide for
the rapid replacement of human resources lost to the infection and the training of additional
personnel to address the urgent issues of prevention, treatment and mitigation.
However, the surge in aid inflows to address the challenges of the HIV/AIDS pandemic may not be
unequivocally good news, and some policy makers and donors fear its possible consequences of
macroeconomic instability and/or Dutch disease effects that may undermine growth. The concern
about the Dutch disease effects is that increased aid will translate into appreciation of the exchange
rate and that this will damage the economy. For instance, when a country receives large inflows of
foreign currency, a significant part of that aid is spent on non-tradable goods, raising their domestic
prices relative to tradable goods prices. The result is a real exchange rate appreciation; which will be
demonstrated mostly through a nominal appreciation in the case of a flexible exchange rate regime
or a rise in domestic inflation in the case of a fixed exchange rate regime. The real appreciation
reduces the competitiveness of the domestic traded goods sectors. Over the long run, production in
these sectors contracts and resources shift to the production of non-tradable goods. This may lead
to a less diversified and more vulnerable economy that is increasingly dependent on external
resource flows. Thus, the short-run welfare benefits of the aid inflows may be superseded by the
welfare losses from the increased cost of non-traded goods and the loss of production in the traded
goods sector. A real exchange rate appreciation also reduces the country’s potential and capacity to
attract investment and grow itself out of poverty and aid dependency27.
Many factors determine the competitiveness of a country’s exports. These include, notably, the
macroeconomic situation, in particular the real exchange rate in the case of a flexible exchange rate
regime; the trade policy regime; the business environment; the cost and availability of credit,
infrastructure; taxes; and so on.
Depreciation of the REER can raise the cost of imported products inducing increased use of local
inputs and savings on imports as agents shift demand to locally produced inputs and goods, while
increasing the profitability of exports. On the other hand, a REER appreciation would reduce the
competitiveness of the domestic traded goods sectors as it lowers returns to entrepreneurial
activity. Over the long run, production in these sectors contract and resources shift to the
production of non-tradables. This may lead to a less diversified and more vulnerable economy that is
increasingly dependent on external resource flows.
Empirical evidence on the effects of the REER on exports in Uganda is, however, very scanty. Atingi-
Ego and Ssebudde (2000), while examining the relationship between misalignment of Uganda’s REER

27
Excellent reviews of the macroeconomic challenges imposed by the scaling up of aid flows are contained in
Gupta, Powell & Young (2006) and Serieux (2006).

91

and non-traditional exports between 1972 and 1999 found that non-traditional exports are
positively related to REER undervaluation and that overvaluation of more than 15% hurts exports.
However, Nkusu (2004a) notes that while the findings of Atingi-Ego and Ssebudde (2000) could
highlight important policy implications, the study could suffer from the omission of weather, which
could have affected agricultural output and thus had a stronger impact on exports than the REER.
Nkusu (2004a) and the Diagnostic Trade Integration Study (DTIS) (2006) also note that besides the
REER, some other factors could be affecting exports, such that an appreciation on the exchange may
not erode export competitiveness. First, Uganda is very likely still producing within its production
possibility frontier. Unused or inefficiently used production factors, such as labour and, to some
extent land, can prevent a resource transfer effect as assumed by the Dutch disease theory. There
are also export (sub-sector) specific issues that affect export competitiveness. In the case of coffee,
improved terms of trade during 1993/94 and1995/96 coupled with the relatively limited REER
appreciation, contributed to large export volumes (Nkusu 2004a). Other factors such as export
diversification and market access issues including high tariffs may also explain the trends in exports.
Besides, appreciation of the REER could be due to an appreciation of the real equilibrium exchange
rate arising from productivity increases, in which case there would not have been an erosion of
competitiveness (DTIS, 2006 pg 18). Biggs (2007) highlights financing costs as export supply
constraints in SSA countries. He notes that in most SSA countries, the relative costs of working
capital credit and fixed asset financing are higher than in competitor countries. This is evidenced by
high REER and high collateral requirements. Access to credit for most borrowers is also limited.
Uganda firms indeed suffer from very low credit market participation and high costs of borrowing
averaging 21% over the review period. The latter appears to be mainly driven by overhead costs,
which in turn are a function of both wages and indirect costs including electricity and
telecommunications (DTIS, 2006 pg 18). Private sector credit in Uganda is significantly lower than the
averages for SSA and comparable low-income countries. Private sector credit to GDP in Uganda was
7.5% in 2004/05. This should be compared to an average of 26% for Kenya and 17% for SSA.
Turning from the relationship between REER and exports to the specific impact of aid flows, how
they affect the real exchange rate and the structure of domestic production, and the size of these
effects, the macroeconomic evidence is weak. Econometric estimates often show these effects to be
small and statistically insignificant.28 However, one problem is that all empirical work in this area is
plagued by severe measurement problems, both of the REER itself and across alternative concepts of
tradable and non-tradable goods.29
Yano and Nugent (1999) find mixed econometric evidence on the relationship between aid, REER,
and the structure of production among a set of 44 aid-dependent countries between 1970 and1990.
For Uganda, they find that although aid is associated with a depreciation of the shilling rather than
appreciation during the period concerned, the non-traded goods sector expanded sufficiently as to
give rise to immiserisation.
Elbadawi (1999) examined the relationship between aid, REER and non-traditional exports for a
sample of 62 countries. He found that a 35% increase in aid levels was associated with a REER
appreciation of 3%. However, exchange rate overvaluation was associated with an increase in non-
traditional exports, rather than the contraction predicted by the Dutch disease model. This also
implied a positive relationship between aid and non-traditional exports. However, that relationship

28
For a summary of the literature see Adam (2006).
29
Adam (2006)

92

was found to be nonlinear. The initial positive relationship eventually becomes negative as aid
increases, exhibiting a Laffer curve effect. The implication is that aid has a positive effect on the
production of tradable goods, but that effect eventually evaporates at very high levels of aid and a
Dutch disease type effect takes over. Elbadawi concludes that Uganda has acute aid dependency and
is likely to experience REER overvaluation.
Sekkat and Varoudakis (2000) examined one aspect of the Dutch disease story for 33 SSA countries
i.e., the relationship between exchange rate overvaluation and manufacturing exports. They found a
negative relationship between exchange rate overvaluation and manufacturing exports. Other
authors (e.g. Rodrik, 2007) have found similar results that REER undervaluation promotes growth
while an overvaluation is a major impediment to growth. However, in the absence of a
corresponding link between aid and overvaluation, this does not amount to a verification of the
Dutch disease story (Serieux, 2007).
Rajan and Subramanian (2005) used both the direct and indirect approaches in investigating the
empirical evidence for aid-related Dutch disease. They examined the relationship between the level
of aid receipts (relative to income) and the performance of exporting sectors versus sectors
producing non-tradable goods. They found that in high aid-receiving countries, the exportable-
producing (tradable) sectors grew significantly more slowly than the sectors producing non-tradable
goods. The authors argue that the Dutch disease explanation for that effect is confirmed by the
finding of a positive relationship between aid and exchange rate overvaluation and between the
retarded growth of export-producing sectors (relative to non-export-producing sectors) in the face
of exchange rate overvaluations. However, Serieux (2007) argues that this is not sufficient to confirm
Dutch disease effects. The observed effects (aid increase, exchange rate appreciation and lower
relative growth of tradable sectors) are also consistent with the condition where there is, initially, an
underutilisation of capacity. Verification of Dutch disease would have to be demonstrated either by
an actual contraction of the export-producing sectors or sustained and substantially slower growth
sufficient to produce a considerable imbalance in the economy over the long run.
The issue of whether increased budget support to finance Uganda’s poverty reduction has resulted
in a Dutch disease has been a subject of much discussion. Neither the behaviour of exchange rates
and performance of exports, nor data of financial flows and macroeconomic performance in general
give any clear cut evidence on the issue.
Nkusu (2004b) argues that the fears for aid-induced Dutch disease in Uganda may be unfounded.
During the period 1992/93 and 1995/96, there was, on an annual average basis, an appreciation of
REER of 7.5%, while the terms of trade improved by 16.6% and total financial inflows increased by
24.3%. Between 1996/97 and 2000/01, the REER depreciated by an annual average of 1%, while the
terms of trade deteriorated by 8% and total financial inflows increased by almost 3%. She argues
that growing financial inflows, developments in terms of trade and structural reforms that the
economy has undergone, indicate that the behaviour of the REER cannot be ascribed to movements
in financial inflows only. Since there has neither been a significant appreciation in Uganda’s REER nor
a decline in real exports, despite massive financial inflows, Nkusu (2004b) concludes that the
applicability of the core Dutch disease model to Uganda has been weak. She specifically advances
three factors that explain the observed weak applicability of the prediction of the core Dutch disease
for Uganda. First, some characteristics of the Ugandan economy depart from key assumptions of the
Dutch disease model, i.e., Uganda is very likely still producing within its production possibility
frontier. Unused or inefficiently used production factors, such as labour and, to some extent land,

93

can prevent a resource transfer effect as assumed by the Dutch disease theory. Second, economic
reforms aimed at liberalising the economy in general and trade system in particular have
encouraged both exports and imports and widened the trade balance. The increase in foreign
inflows compensated for the shortfall in exports proceeds emanating from the adverse terms of
trade shock that has affected Uganda since 1998/99. And finally, prudent monetary and exchange
rate management have achieved price stability and has controlled real exchange rate appreciation.
However, she acknowledges that there is a limit to the level of aid that can be managed, beyond
which it could exceed the sterilisation capacity of the monetary authorities and render
macroeconomic management difficult and even undermine the growth prospects. Developments
since the period considered may be an illustration of this, with even larger aid inflows. Atingi-Ego
(2005) notes that more recent appreciation pressures should at least be partly attributed to the
strong sterilisation effects on account of a shilling injections resulting from the donor flow to finance
government fiscal deficits.
In recent years, simulation models calibrated by data have been used to understand the dynamic
responses and assess the quantitative significance of the macroeconomic effects of aid flows.
Adams and Bevan (2003) develop a Computable General Equilibrium (CGE) model of aid and public
expenditure where public infrastructure capital generates inter-temporal productivity spillover for
both tradable and non-tradable sectors. The model also provides for a learning-by-doing externality,
through which total factor productivity in the tradable sector is an increasing function of past export
volumes. The model is calibrated to contemporary conditions in Uganda to simulate the effect of
increased aid. They find that public expenditures whose productivity effects are skewed towards the
non-tradable sector deliver the highest growth in exports and total output. The bias in productivity
effects increases the supply of non-tradable goods, which is sufficiently strong to almost entirely
offset the demand effects of increased aid flows. The results also show that exchange rate
appreciation is reduced or even reversed enhancing export sector performance. However, in terms
of poverty reduction, the results show that income gains largely accrue to urban skilled and unskilled
households leaving the rural poor relatively worse off.
In a discussion paper, Atingi-Ego (2005) comments on the studies by Nkusu (2004) and Adam and
Bevan (2003), that both point to the fact that the fears for a Dutch disease in Uganda may be
unfounded. Whereas he acknowledges the Nkusu argument - that if there is excess capacity (which
could be the case in Uganda), then an increase in aid will only move the economy closer to the
production possibility frontier - he is more skeptical to the assumptions encompassed in the Adam-
Bevan (2003) model. First, for the productivity spill-over effects to counteract the appreciation
tendency, the price elasticity of supply must be elastic, i.e. larger than one. This is most likely not the
case in Uganda, at least in the short run. Second, he questions the extent of these spill-over effects.
Uganda lacks institutions to ensure efficiency of the investments, i.e. value for money. Atingi-Ego
claims that even if Dutch disease exists in Uganda, the impact cannot be distinguished. Since exports
and non-traditional exports in particular, are increasing, there are probably certain sub-sectors in the
export sector that could be realising productivity efficiencies to be able to offset appreciating export
rates.
Examining the trends in the price indices for the major components of GDP, Atingi-Ego finds that
prices for non-traded goods in Uganda have grown much faster than prices for traded goods, exactly
as Dutch disease theory would predict. This implies a shift of price incentives away from the
production of traded goods towards non-traded goods in the last few years on account of the

94

increased demand for non-traded goods arising from increased government expenditures. Given the
fixed supply of these goods in the short-run, price increases have been the inevitable. It is also
possible that the aid-funded fiscal expansion has contributed to the increase in the trade deficit from
7.2% of GDP in 1997/98 to 10.1% of GDP in 2003/04. A shift in relative prices from tradable to non-
tradable goods might also undermine the national objective of creating a dynamic export-led
economy. Private sector–led export promotion is central to the Medium Term Competitiveness
Strategy (MTCS) and this objective should not be compromised by an excessive fiscal deficit.
A joint study carried out by the Ministry of Finance, Planning and Economic Development (MFPED)
and the BoU in (2005) to investigate the impact of the exchange rate appreciation observed during
2003/04 on competitiveness of the export sector found general negative consequences of the
appreciation on some exports namely: (i) Reduction in export profitability (and even large losses in
some sectors) and/ or reduction in farm gate prices, with reduced incentives having major
implications for future production and value addition; (ii) Reduced export competitiveness and loss
of major contracts to foreign competitors; and (iii) Reduced investment in the export sector. The
paper demonstrates that the appreciation of the shilling that has occurred in recent years has
adversely affected exporter profitability and export competitiveness and may have contributed to an
increase in poverty, and that “Uganda’s large aid-financed fiscal deficit is not necessarily compatible
with the objectives of poverty reduction and strong export-led growth, even though its purpose is to
finance expenditures aimed at poverty reduction” (MoFPED-BoU, 2005, p.37).
Atingi-Ego (2005) notes that these short-term consequences also have significant longer-term
implications for the wider economy in that a permanently appreciated REER will discourage export
diversification and export-led growth in general, in addition to shifting incentives towards the non-
tradable sector and encouraging imports. Lower farm-gate prices also reduce rural incomes and thus
reduce demand for locally manufactured goods and services. The increase in poverty between 2000
and 2003 is partly attributed to falling farm-gate prices of several export crops and coincides with a
slowdown in formal manufacturing growth.
Other macroeconomic consequences of aid observed in Uganda include: (i) High fiscal deficits; (ii)
High domestic interest costs arising out of sales of Treasury bills to sterilise shilling liquidity
injections from aid induced government spending; and (iii) High real and nominal interest rates.
Alongside the REER, exchange rate volatility, which most often is salient in flexible exchange rate
regimes, can hurt exports. Friedman (1953) and Johnson (1969), however, argue that flexible
exchange rates could be beneficial to the economy as they promote trade and overall
macroeconomic stability. Other researchers have maintained that short-run fluctuations in exchange
rates have no effects on trade. For instance, Bailey, Talvas and Ulan (1987) point out that traders
anticipate future exchange rate movements better than the average exchange rate participant and
gains from this knowledge could offset the risk of exchange rate uncertainty. But, Bailey et al
continue, if the exchange rate volatility is due to fundamentals, efforts by authorities to reduce
volatility by means of controls or other restrictions on trade could be more harmful to trade and
could reduce it.
Some researchers, for instance, Kihangire et al (2005), Mundell (2000), Doroodian (1999) and
Krugman (1989) argue that because of market imperfections, particularly in developing countries,
hedging against exchange rate volatility is both imperfect and very costly as a basis for avoiding
exchange rate risk. Therefore, exports may be negatively affected by exchange rate volatility.
Kihangire et al (2005) highlight several reasons that explain why exports can be affected by exchange

95

rate volatility: (i) Most export contracts in Uganda are priced and paid for in foreign currency,
therefore exchange rate variability affect export earnings valued in domestic currency. (ii) Export
contracts may involve long time lags due to production delays, delivery lags and the actual
settlement date, all of which may increase the extent of uncertainty. (iii) Imperfections regarding
hedging facilities may make it difficult to fully anticipate and contain uncertainty caused by exchange
rate variability and the extent of export product diversification and market power determine a firm’s
ability to suffer or export the risk. Kihangire et al (2005) in an investigation of the effects of exchange
rate volatility on flower exports between 1994 and 2001 found that a negative relationship between
exchange rate volatility and exports of flowers existed.
Overall, the evidence on the determinants on export competitiveness in Uganda is inconclusive and
requires further empirical investigation especially in view of the observed appreciating REER in the
recent past and its possible effects on the economy. Besides, export (sub-sector) specific issues that
affect export competitiveness underscore the need to investigate (sub-sector) specific determinants
of exports. This study examines six exports: coffee, tea, cotton, fish, maize and flowers.

3. Descriptive Analysis
Macroeconomic stability, coupled with continued GoU commitment to reforms and increased
confidence among external donors, has attracted a wave of increased official and private aid flows to
Uganda. Over the last decade, official aid transfers (grants to government) averaged close to US$
500 million per year, or 7% of GDP. In recent years, private aid transfers (to non-government
recipients) have risen sharply, to comprise more than 50% of total aid receipts in 2005/06 and
2006/07 (see Figure 42). Atingi-Ego (2005) notes that the increase was as a result of the
development of the first PEAP in 1997/98, which led to the introduction of the PAF in 1998/99 and
the eventual qualification of Uganda for the HIPC initiative in 1998/99. Much of the official donor
inflows have been spent on poverty reduction programmes including health, and primary education,
water and sanitation, whose share in the PAF averaged about 77% over the decade. The share of the
Health sector in PAF averaged 17%. In addition the health sector has been receiving increasing
amounts of aid, mainly as project funding towards the fight against HIV/AIDS. The share of the
Health sector in donor projects, for instance, averaged about 18% over the last five years. The share
of total aid flows (to both government and non-government recipients) related to HIV/AIDS has been
relatively small, but is rising sharply, from an estimated 4% in 2003/04 to 16% in 2006/07.
Aid has enabled the country to continue growing at an impressive rate, improvements in social
welfare and, thus, a reduction in poverty. Over the last decade, GDP averaged about 6%, while the
level of poverty declined from 44% in 1997/98 to 31% in 2005/06. Interventions30 in the fight against
HIV/AIDS have led to a reduction in prevalence from a peak of 18% in 1992 to the current figure of
6%.
Whereas the surge in aid inflows carries good news regarding prospects for improving welfare in the
country, it has raised a number of macroeconomic concerns ranging from fears about its possible
effects on REER and export competitiveness, inflation, fiscal performance and debt sustainability (for
non-grant components of aid). This chapter focuses on the first two effects: real exchange rates and
inflation.

30
Interventions included information, education and communication (IEC), laboratory and blood transfusion
services, sexually transmitted diseases management, care and support.

96

Figure 42: External Aid Transfers (Grants) to Uganda, 1995/06 - 2006/07


Liquidity Injection31 and Money Growth
One major concern about the surge in aid inflows is that aid-induced spending by government tends
to increase money supply in the domestic economy. However, as Figure 43 indicates, liquidity
injections caused by government expenditure were not always associated with increases in money
supply, except for the period 1998/99 to 2002/03. On the one hand, liquidity injections may not
necessarily have translated into inflationary and REER appreciation pressures basing on a number of
factors: first, prudent monetary policy, which enabled liquidity injections to be sterilised; second, a
large part of aid was used for government imports mainly for projects. Government project imports
averaged US$ 106.52 million, compared to non-project government imports, which averaged US$
36.44 million over the period 1993/94 to 2006/07. Further, as noted in Chapter 4 of this report,
about 60% of aid towards supporting HIV/AIDs was spent externally, mainly on the importation of
drugs. Third, spare capacity such as unused land, labour, and capital in the economy means that any
increased government spending on domestically produced non-tradable goods can be met by
adequate supply response, which in turn neutralises the Dutch disease effect. On the other hand, it
is plausible to think that liquidity injections could have led to volatility in money supply, which could
lead to volatility in prices and exchange rates in the short run, since anyway, injections are sterilised
by the monetary authorities with a lag. Moreover, aid may impose other costs such as high domestic
debt service burden and crowding out of private sector investment. This study empirically
complements these arguments in order to draw a firm conclusion regarding possible movements in
the REER and inflation that may be caused by aid.

31
Calculated as total government domestic expenditure less total domestic revenues.

97

Figure 43: Liquidity Injection and Money Supply Growth


Sterilisation and its Costs
The main instrument used to sterilise the liquidity injections arising from aid spending by the
monetary authorities was sales of government securities (Treasury bills and bonds). Total holdings of
securities by the private sector stood at Shs 2,293.4 billion in June 2007 up from Shs 50.03 billion in
June 1999. The associated interest cost to government was Shs 181.1 billion in June 2007, up from
Shs 1.6 billion in June 1999 (Table 24). High domestic interest costs to the government would put
pressure on the fiscal balance excluding grants, and raise concerns about medium-term fiscal
sustainability and domestic debt sustainability, especially in a shallow financial market. The fiscal
deficit as a percentage of GDP rose from 6% in 1997/98 to a peak of 13% in 2001/02; although it
improved to 9% in 2006/07, it remained higher than the level of 6% recorded in 1997/98. Besides, a
crowding out of banks lending to the private sector and hence private investment, can result from
sale of Treasury bills and bonds since commercial banks would find it risk free and more profitable to
invest in those securities than lend to the private sector. The relatively high Treasury bills rates in the
region has also attracted portfolio investment inflows in the domestic money markets exerting
nominal appreciation pressures on the exchange rates.
Table 24: Government Securities and Interest Costs (billion shillings) (as at June)

(June) Face Value Cost Total Interest Interest cost
(%) of total security
holding
1999 50.03 48.48 1.55 3.09
2000 81.41 76.91 4.50 5.53
2001 68.10 62.69 5.41 7.95
2002 115.54 107.63 7.91 6.85
2003 131.53 114.44 17.08 12.99
2004 169.56 156.71 26.05 15.36
2005 83.88 76.19 9.69 11.55

98

32
2006 1,845.92 1,695.50 150.43 8.15
2007 2,293.43 2,112.31 181.12 7.90

Broad Money (M3) Growth and Inflation in Uganda
Overall, broad money (M3) growth indicated considerable decline together with headline inflation
rates over the period under review, although the relationship between money and underlying
inflation is less clear, particularly over the last decade. Although there were indications of counter-
cyclical relationships during 2001/02 and 2004/05, overall, there seemed to be a positive
relationship between broad money growth and headline inflation in Uganda.
Figure 44: Inflation and Broad Money (M3) Growth in Uganda


Inflation and Real Effective Exchange Rates
Higher inflation would be expected to cause the REER to appreciate (decrease), unless offset by
rising inflation in trading partner countries or a depreciation (increase) in the nominal exchange rate.
As noted in Figure 45, changes in the REER were primarily driven by changes in the NEER rather than
inflation; NEER depreciation was generally sufficient to offset higher inflation and cause the REER to
depreciate even when inflation was rising. Only in a few instances (e.g. in 2004/05 and 2006/07 was
there a combination of rising inflation and nominal appreciation that caused the REER to appreciate.
Overall, this descriptive analysis is not conclusive regarding aid-induced effects on inflation and
REER.

32
The drastic rise in Treasury bill/bond sales observed in 2006 to Shs 1,846 billion from Shs 84 billion in 2005
was caused by the need to sterilise excess liquidity that resulted from positive developments in the balance
of payments, which increased the demand for Shillings.

99

Figure 45: Nominal and Real Effective Exchange Rates and Inflation (annual averages)


Real Effective Exchange Rate and Export Receipts
Figure 47 through to Figure 53 indicate the movements in the REER and/or terms of trade (TOT), and
total export receipts in values as well as volumes of coffee, cotton, tea, fish, maize and flowers,
which together accounted for 44% of Uganda’s export receipts in 2005/06 (see Table 25). The three
agricultural export crops i.e. coffee, cotton, and tea—analysed in this study made up around 20% of
Uganda’s merchandise export receipts in 2005/06, and were the main sources of income for around
17% of the population in 2004/05 (DTIS, 2006). Of these three, coffee is by far the most important,
contributing around 17% of the total merchandise export revenues in 2005/06, and being the main
source of income for 11% of the population. This is followed by tea (2.5%) and cotton (1%) of
merchandise exports. While coffee exports had fallen significantly from the peak reached in the mid-
1990s, cotton and tea exports had been fluctuating around an upward trend, although their share in
total export revenues declined to 1,2% and 2.5% in 2005/06 from 6.4% and 5.9%, respectively in
2003/04.
The differing performances of these three sub-sectors reflect, in part, differing trends in their
international prices. Although coffee prices have been rising in the last four years, they are still way
below the peaks reached in the mid-1990s, while cotton and tea prices have been fluctuating around
a relatively flat trend over the last decade. More importantly, they reflect crop-specific issues, which
if addressed would increase their production and exports even in the context of long-term declines
in international commodity prices (DTIS, 2006).
Fish exports have grown rapidly over the last decade to become the largest merchandise export item
in 2005/06, with an 18.5% share. The fish sector is very important for poverty reduction: it is the
main source of income for some 266,000 households, equivalent to around 1.2 million people or 4%
of the population. Maize is another non-traditional export, which is particularly important for
regional trade. It accounted for about 2% of total export revenues in 2005/06. Flower exports,
whose major destination is the European Union (EU) increased in importance over the last decade,

100

with their contribution to total export earnings rising from 1.5% in 1997/98 to 3% in 2005/06. The
increasing importance of floriculture is reflected in the fact that the sector employs about 6,000
people (DTIS, 2006).
Table 25: Selected Exports Receipts (million US$)

Coffee Cotton Tea Fish Maize Flowers Sub- Total % of total
total exports exports
92/93 111.3 5.3 10.0 4.4 7.9 0.0 139.0 172.7 80.4
93/94 180.0 4.3 12.1 11.1 14.0 0.0 221.5 264.7 83.7
94/95 456.6 3.3 9.4 17.0 20.1 2.0 508.5 593.0 85.8
95/96 404.4 13.3 10.0 37.6 9.4 5.4 480.0 585.5 82.0
96/97 365.6 28.6 21.3 34.6 16.5 5.3 472.0 682.7 69.1
97/98 268.9 11.4 35.0 28.0 8.1 6.8 358.2 458.4 78.1
98/99 306.7 10.8 22.7 47.6 5.9 7.2 400.9 549.1 73.0
99/00 186.9 22.5 31.9 24.8 4.0 8.3 278.3 459.9 60.5
00/01 109.6 14.1 35.9 66.6 6.1 13.2 245.7 458.3 53.6
01/02 85.3 18.0 26.9 107.5 13.1 15.9 266.6 474.0 56.2
02/03 105.5 16.9 29.5 111.4 8.2 17.0 288.4 508.5 56.7
03/04 114.1 42.8 39.3 118.1 18.8 27.2 360.3 670.9 53.7
04/05 144.5 41.3 33.1 169.6 13.3 31.7 433.6 886.3 48.9
05/06 173.4 12.9 25.6 192.8 23.7 32.7 461.0 1,042.5 44.2
Table 26: Share of Total Exports

Coffee Cotton Tea Fish Maize Flowers Total
1992/93 64.4 3.1 5.8 2.6 4.6 0.0 80.4
1993/94 68.0 1.6 4.6 4.2 5.3 0.0 83.7
1994/95 77.0 0.6 1.6 2.9 3.4 0.3 85.8
1995/96 69.1 2.3 1.7 6.4 1.6 0.9 82.0
1996/97 53.6 4.2 3.1 5.1 2.4 0.8 69.1
1997/98 58.7 2.5 7.6 6.1 1.8 1.5 78.1
1998/99 55.9 2.0 4.1 8.7 1.1 1.3 73.0
1999/00 40.6 4.9 6.9 5.4 0.9 1.8 60.5
2000/01 23.9 3.1 7.8 14.5 1.3 2.9 53.6
2001/02 18.0 3.8 5.7 22.7 2.8 3.4 56.2
2002/03 20.7 3.3 5.8 21.9 1.6 3.4 56.7
2003/04 17.0 6.4 5.9 17.6 2.8 4.0 53.7
2004/05 16.3 4.7 3.7 19.1 1.5 3.6 48.9
2005/06 16.6 1.2 2.5 18.5 2.3 3.1 44.2

It is observed that the REER had an impact on Uganda’s exports during some periods, but the
relationship seems the opposite of what would be expected in other periods, and varied across
different exports. Figure 46 shows that between the periods 1992/93 and 1996/97, while the REER
appreciated steadily, export receipts increased rapidly, probably on account of the 1993/94 coffee
boom as a result of the rise in international coffee price and the subsequent TOT improvement.
However, between 1997/98 and 2000/01, although the REER depreciated, export receipts exhibited
both periods of decreases and increases, suggesting a mixed REER influence on exports. Between
2002/03 and 2006/07, a sharp increase in the total export receipts was preceded by a depreciation
of the REER. However, total export receipts seemed to have moved in line with TOT (see Figure 47).
For example, the sharp improvement in TOT of the mid 1990s, which was due to the rise in

101

international prices of coffee, was followed by a sharp increase in export receipts. Deterioration in
TOT in the late 1990s and early 2000s was followed by a fall in export receipts. In the past three
years, when TOT showed an improvement in export receipts.
Figure 46: Evolution of the REER and Total Export Receipts

Figure 47: Total Exports (US$ mn) and Terms of Trade

Examination of the relationship between the REER and specific exports further confirms the
inconsistent relationship between the REER and exports and also reveals that the relationship varied
by type of exports (see Figure 48 through to Figure 53). In the case of coffee (Figure 48), a sharp rise
in export volumes is noted during the period 1995/96 to 1997/98, when the REER was appreciating.
This could mainly be explained by a rise in the international prices of coffee during that time. For
much of the last five years, however, coffee exports have been declining although the REER was
either rising or relatively constant. As noted by the DTIS (2006), the main issue with coffee is the

102

coffee wilt disease (CWD), which has affected more than half of the robusta trees (over 80% of
Uganda’s coffee is robusta). The loss due to CWD is around 44% of the estimated 2005/06 output,
and around one-quarter of recent export revenues. Figure 49 and Figure 51 indicate that tea and fish
moved fairly in line with the REER over the periods 1997/98 to 2004/05 and 1999/2000 to 2004/05,
respectively. Cotton and maize showed the most contrasting relationships between export volumes
and the REER over the review period (see Figure 50 Figure 52). However, fish and flower exports
moved quite in line with the REER over the review period (see Figure 51 and Figure 53).
This analysis not only points to the fact that the relationship between the REER and exports might be
weak for some exports and uneven across different exports in Uganda, but also that there could be
other factors (besides the REER) affecting exports. Nkusu (2004) for example noted that Uganda
being a predominantly agricultural country, adverse weather conditions that had affected the
country repeatedly could have had a more stronger impact on exports than even the REER and TOT
shocks. Deininger and Okidi (2001) counted weather and diseases among the common determinants
of changes in yields for many crops in Uganda between the period 1992 and 1999.
Figure 48: Coffee Exports (volume) and REER

103

Figure 49: Tea Exports (volume) and REER

Figure 50: Cotton Exports (volume) and REER

104

Figure 51: Fish Exports (volume) and REER

Figure 52: Maize Exports (volume) and REER

105

Figure 53: Flower Exports (value, US$ million) and REER

4. Estimation Methods and Data
Two different models are estimated in this chapter: one which links aid flows, inflation and the REER,
and another which links the REER and exports. Details of the models are given at Appendix 1. The
first uses a Vector Autoregressive (VAR) analysis to derive the responses of REER and inflation to aid
flows. The second comprises an export supply model that assumes supply side constraints as the
major factors affecting exports, including the REER, costs and availability of working capital.
The data used in the empirical analysis of aid, inflation and the real exchange rate is a monthly time
series, which covers the period July 1994 to June 2007. All data series are transformed to logarithmic
form. The series on macroeconomic variables, i.e. aid (FTA), REER, money supply (BM3) and prices
were obtained from quarterly and annual reports published by BoU. We use two measures of prices;
underlying and headline consumer price index (CPI). Underlying CPI excludes food prices while
headline CPI includes prices of all items. Money supply (BM3) includes both shilling and foreign
currency deposits. The trade weighted REER rebased (100=2000) were obtained from the BoU
database. The REER for Uganda is computed as the NEER index adjusted for relative movements in
national prices and its trading partners. A decrease in the REER implies an appreciation whereas an
33
increase implies depreciation .
For the export model, quarterly data for the period 1993:Q3 to 2006:Q4 is used. All data series
except export unit prices and the lending rate are transformed to logarithmic form and are
seasonally adjusted unless otherwise stated. The NEER (Shs/US$) and trade weighted REER rebased
(2000 = 100) were obtained from the BoU database. The REER is computed as above. Exports in
million 60-kg bags of coffee and in thousand metric tonnes of tea, tobacco, fish and maize and their
unit prices in US$ per kilogram were obtained from the BoU database. The value, rather than the
volume, of flower exports is used because data on the price of flowers is not available on continuous
basis. The selection of commodity exports was based on their contributions to total export volume.

33
For further details on the computation of the REER see Appendix Table 8.

106

The total contribution of all the exports studied was 44% in 2005/06. Rainfall figures are measured in
millimetres and were obtained from Statistical Abstracts published by the UBOS. The cost of capital
Lxt is the quarterly average lending rate. Exchange rate volatility (VAVB) is computed as a quarterly
average of volatility obtained from a GARCH (1, 1) model for the monthly nominal exchange rate
series (NER).

5. Estimation Results
Identifying the Effects of Aid
The VAR results (provided in detail in Appendix 2) show that aid flows have an impact on the money
supply but not on inflation, perhaps reflecting that due to sterilisation measures of liquidity by the
monetary authorities, injections of liquidity into the economy following government expenditure
would not be inflationary, though they would lead to short run increases in money supply. This is
further reflected by the fact that money supply growth does not cause higher prices. However, both
aid and money are associated with changes in the REER implying a possible aid induced effect on the
REER, although conclusions cannot be established at this level until we examine the impulse
response functions.
The next step therefore captures the responses of the money supply, CPI and REER to an increase in
aid using impulse response functions. Figure 54 and Figure 55 display the responses of the other
variables in the system to shocks to aid, money supply, CPI and the REER, over a horizon of 12
months following a shock, using headline and underlying inflation, respectively.
Focusing on column 1 of Figure 54 which indicates the response of the macroeconomic variables to
shocks to aid, it can be seen that changes in aid are associated with volatility in all the other
variables: money supply (column 1, figure 2), CPIH (column 1, figure 3) and REER (column 1, figure
4). Specifically a one standard shock to aid would be followed by an initial increase in money supply
of 0.05 standard deviations after about 1 month, and then followed by fluctuations up and down in
money supply within 0.05 – 0.1 standard deviations. The fluctuations of prices (up and down) to a
change in aid start at about the fourth month with 0.15 standard deviations. They are always
negative and do not exceed 0.15 standard deviations. The last figure in column 1 indicates that an
increase in aid leads to volatility in the REER over the next year. In about the second month
following a 1 standard deviation increase in aid, the REER depreciates by about 0.05 standard
deviations, it then fluctuates down and up over the coming periods within a range of 0.05 to – 0.075
standard deviations. However, although these findings indicate that aid would lead to
macroeconomic instability through volatility in prices and REER, the impact is very small.
Consistent with the results in Figure 54, the second last figure of column 2 shows that money supply
hardly causes any changes in prices, until after one year. This could be explained by the fact that
injections of liquidity brought about by government expenditure would be sterilised through
monetary policy (for instance through sales of Treasury bills). The last figure in column 2 indicates
that an increase in money supply which could result from government expenditure would be
associated with both depreciations and appreciations (volatility) of the REER starting with
depreciation in the second to fifth month. The magnitudes of the changes are very small. The last
figure in column 3 indicates that changes in prices are associated with very small increases in the
REER starting in the first month, and later smaller decreases in the REER starting in the sixth month.
Alhough these results would indicate possible macroeconomic distortions caused by aid on the
economy via volatility in key prices (inflation and REER), the impact is very small.

107

Similar results are obtained when the underlying CPI is used, with the exception that the rise in
underlying CPI starts early in the second month following an increase of aid. Underlying CPI then
starts falling after the forth month; thereafter the effect disappears (see the second last chart,
column 2 of Figure 55).
This study specifically finds that although aid may have an adverse impact on the economy through
causing short-run volatility in prices and REER, the impact is small. This could be explained by the
fact that there are measures by the monetary authorities to sterilise liquidity injections arising from
aid. However, volatility in the REER may be damaging to private sector investment as it would lead
to temporary losses to exporters. Indeed, two studies on the Uganda economy found that exchange
rate volatility discouraged the export of some goods (see Kihangire et al, 2005; Nannyonjo and Apaa,
2007). Moreover, aid dependence could also distort institutional development; in this case through
high transaction costs by requiring substantial resources to be allocated to the coordination and
management of aid flows, particularly in an economy like Uganda where financial markets are thin.
As Table 24 indicates, management of aid by the monetary authorities was associated with interest
costs of the Treasury bills averaging 8% of the total stock of Treasury bills over much of the review
period. High domestic interest costs to the government would put pressure on the fiscal balance
excluding grants and carry negative implications for medium-term fiscal and domestic debt
sustainability, especially in a shallow financial market. Moreover, the fact that the largest proportion
of government securities is held by the commercial banking sector could lead to crowding out of
private sector development.

Real Exchange Rate and Exports
The results of the modeling exercise34 indicate that the REER affects some Uganda exports, in line
with the predictions of economic theory that a depreciation or appreciation of the REER would
encourage or discourage production of exports, in particular fish, flowers and cotton. The other
variables that affect exports are unit prices/TOT, rainfall and the lending rate although the effect
varies by type of export.
Table 27 and Tables A7-A13 (Appendix 2) indicate that the main factors affecting total exports in
Uganda are TOT and volatility of the exchange rate. The coefficient on the TOT variable is positive
and significant, indicating that an improvement in TOT would encourage exports. Specifically a 1%
improvement in TOT would lead to an increase in exports receipts of 0.6%. Though the coefficient on
VAVB is significantly positive, it is not consistent with the fact that exchange volatility would create a
risky business environment that would discourage exports. We do not establish a significant
relationship between the REER and total exports. However, this was not the case for all the
individual export commodities as explained below:
Table A8 indicates that exports of coffee are mainly affected by rainfall and the cost of capital. In
particular, a 1% increase in rainfall leads to a 0.8% increase in coffee exports. An increase in the
lending rate of 1% reduces coffee exports by 0.6%.
Table A9 indicates that tea exports mainly adjust to deviations from long run equilibrium. This
implies that any shock to tea export production that drifts it away from its long run path would not
last for a long period of time.

34
An error correction model was used. See Appendix 2 for detailed results.

108

Table A10 indicates that cotton exports are affected by the REER, although the effect is very small. In
particular, an increase in the REER (depreciation) is associated with an increase of 0.04% in exports.
The exports of cotton are also negatively affected by their past values.
Table A11 indicates that a depreciation of the exchange rate leads to a 2.7% increase in fish exports.
This finding would support the view that an appreciation of the REER would hurt export
competitiveness. It would also have negative implications for poverty reduction, given that fish is a
rapidly growing export (constituting one of the largest merchandise export item) and being the main
source of income for some 266,000 households, equivalent to around 1.2 million people or 4% of the
population. This finding supports the earlier findings by Atingi-Ego and Ssebudde (2000). Finally, fish
export is negatively related to its past values.
Table A12 indicates that the main factors affecting maize export are rainfall and its past values. The
coefficients on rainfall and maize past values are positive and statistically significant. Specifically, a
1% increase in rainfall would help increase maize export supply by 0.5%.
Table A13 indicates that the main factors affecting flower exports are the REER and cost of capital.
Specifically, a 1% increase in the REER would help increase supply of flower export by 3%. An
increase in the cost of capital by 1% would lead to a fall in flower exports by 1.5%. Rainfall and time
(technological change) also affected the export of flowers. More rainfall discouraged the supply of
flowers. In particular, a 1% increase in rainfall reduced supply of flower exports by 0.3%. Flower
exports were also positively affected by developments that occurred over the review period as
reflected by a positive and statistically significant coefficient on the time trend variable.
Table 27: Responsiveness of Exports to 1% Change in Independent Variable

Exports Terms of Trade Real Exchange Rainfall Exchange rate Interest rate
(ToT) Rate (REER) volatility
Total Exports 0.6 .. .. .. ..
Coffee .. .. 0.8 .. -0.6
Tea .. .. .. .. ..
Cotton .. 0.04 .. .. ..
Fish .. 2.7 .. .. ..
Maize .. .. 0.5 .. ..
Flowers .. 2.89 -0.33 .. -1.5
.. no effect or not significant

Source: Appendix 2, Tables A7-A13

6. Conclusions and Policy Recommendations
This chapter analyses the impact of aid flows to the Ugandan economy on prices and REER over the
period July 1994 – June 2007, using a VAR. It finds that an increase in aid flows is associated with a
long-term increase in the money supply, as expected. However, this does not lead to any long-term
increase in prices or to REER appreciation, which suggests that the BOU’s monetary policy and
sterilisation strategy has been successful. In the short run, an increase in aid is associated with
greater volatility in both prices and the REER, although the impact is small. Volatility in the REER
could be damaging to private sector investment. Moreover, aid dependence leads to high interest
costs by requiring substantial resources to be allocated to the management of aid flows by the

109

monetary authorities. This would have negative implications for medium-term fiscal sustainability
and domestic debt sustainability.
The chapter also attempts to quantify the relationship between the REER and exports in order to
provide insights about the possibility of an aid induced Dutch disease effect on the Ugandan
economy that could partly be caused by aid towards the control of HIV/AIDS. To accomplish this
objective, a model of determinants of six Ugandan exports (coffee, tea, cotton, fish, maize and
flowers) is estimated using vector error correction analysis and quarterly data over the period 1993-
2006. The results indicate that the REER would affect specific exports, namely, fish, flowers and
cotton, which together account for nearly a quarter of Uganda’s exports. Thus, for fish, flowers and
cotton, the findings indicate a possible Dutch disease effect; although firm conclusions cannot be
made unless the study is complemented by an examination of the link between aid and the REER.
This is an issue that is left for future research. Moreover, the study could have benefited from
including additional explanatory variables, which are likely to constrain Uganda’s exports, including
infrastructure such as electricity, road network, among others. To the extent that a possible Dutch
disease effect would reduce supply of some exports, poverty reduction efforts in the long run would
be negatively affected given that a significant proportion of the rural households are dependent on
these export commodities. This underscores the need to contain appreciation pressures that may
arise from aid flows, which have played a big role in Uganda’s poverty reduction programmes over
the last decade, in particular control of HIV/AIDS. Finally, the findings indicate that weather (rainfall),
unit prices/TOT, and cost of capital also affect Uganda’s exports although to varying degrees
depending on the type of export.
Figure 54: Identifying the Effects of Aid using Headline Inflation

LTFA (LTFA eqn) LBM3 (LTFA eqn)
0.000 LCPIH (LTFA eqn) LREER (LTFA eqn)
0.50 0.00 0.005
-0.002
0.25
-0.01 0.000
0.00 -0.004
0 5 10 0 5 10 0 5 10 0 5 10
LTFA (LBM3 eqn)
0.04 LBM3 (LBM3 eqn)
0.000 LCPIH (LBM3 eqn) LREER (LBM3 eqn)
0.1 0.015
0.03
-0.002 0.010
0.0 0.02
-0.004 0.005
0.01
0 5 10 0 5 10 0 5 10 0 5 10
LTFA (LCPIH eqn) LBM3 (LCPIH eqn) LCPIH (LCPIH eqn) LREER (LCPIH eqn)
0.00 0.015 0.012 0.005
-0.05 0.010
-0.10 0.005 0.010 0.000
-0.15 0.000
0 5 10 0 5 10 0 5 10 0 5 10
LTFA (LREER eqn)
0.010 LBM3 (LREER eqn)
0.0075 LCPIH (LREER eqn) LREER (LREER eqn)
0.05 0.020
0.005 0.0050
0.00 0.015
0.000 0.0025 0.010
-0.05
0 5 10 0 5 10 0 5 10 0 5 10

110

Figure 55: Identifying the Effects of Aid using Underlying Inflation

LTFA (LTFA eqn)
0.01 LBM3 (LTFA eqn) LCPIU (LTFA eqn) LREER (LTFA eqn)
0.50 0.000 0.0000
0.00
0.25 -0.001 -0.0025
-0.01
0.00 -0.002
-0.02 -0.0050
0 5 10 0 5 10 0 5 10 0 5 10
LTFA (LBM3 eqn) LBM3 (LBM3 eqn) LCPIU (LBM3 eqn)
0.010 LREER (LBM3 eqn)
0.10 0.03 0.001
0.05 0.02 0.005
0.00 0.000
0.01 0.000
-0.05
0 5 10 0 5 10 0 5 10 0 5 10
0.1 LTFA (LCPIU eqn)
0.015 LBM3 (LCPIU eqn)
0.0070 LCPIU (LCPIU eqn)
0.0000 LREER (LCPIU eqn)

0.0 0.010 0.0065 -0.0025
0.005 0.0060 -0.0050
-0.1
0.000 0.0055 -0.0075
0 5 10 0 5 10 0 5 10 0 5 10
LTFA (LREER eqn)
0.010 LBM3 (LREER eqn) LCPIU (LREER eqn)
0.020 LREER (LREER eqn)
0.05 0.0050
0.005 0.015
0.00
0.0025 0.010
-0.05 0.000
0.005
0 5 10 0 5 10 0 5 10 0 5 10

111

Appendices
Appendix 1: Models and Estimation Methods
Aid flows, inflation and the real exchange rate
Vector Autoregressive analysis is used to derive the responses of REER and inflation to aid flows.
Assume that aid affects inflation and REER via money supply. Let the time path of inflation (CPI),
REER and money supply (BM3) be affected by current and past realisations of the aid (F) sequence.
The transmission of aid to the economy can be represented by the following structural model:

M t = α M + β MM M t + β MF Ft + AM ( L) X t + υ Mt
Ft = α F + β FM M t + β FF Ft + AF ( L) X t + υ Ft (1)

where αM and αF are vectors of intercept terms; Mt is a vector of macroeconomic variables (including
REER, prices and money supply); Ft is a vector containing aid variables; βij are matrices of impact
35
multipliers; AM(L) and AF(L) are kth-order matrices of structural polynomials in the lag operator L
(such that B(L) = B1L + B2L2 +…+ BkLk); Xt = [Mt´ Ft´]´; and υMt and υFt are vectors of structural
(orthogonal) disturbances. The first equation in (1) describes the behavioural relationships between
Mt and all variables in the VAR model, including the aid variables Ft. The second equation in (1)
describes the relationship between the aid variables (Ft) and the macroeconomic variables.
In order to estimate the system, it must be transformed into a usable form, or the VAR in standard
form:
M t = AM + BM ( L) X t + e Mt
Ft = AF + BF ( L) X t + e Ft (2)

where AM and AF are vectors of intercept terms; BM (L) and BF (L) are kth-order matrix of polynomials
in the lag operator L; and eMt and eFt are vectors of structural disturbances. It is assumed that eMt and
eFt are serially uncorrelated with constant variances.
Since the right hand side of the equations in (2) contains only predetermined variables and these are
by definition not correlated with the error terms (assuming no serial correlation), consistent
estimates can be obtained from each equation using Ordinary Least Squares (OLS). However,
estimating the VAR requires an appropriate lag length to be determined. This is done by imposing
cross-equation restrictions that reduce the number of lags. Although a likelihood ratio test would be
applicable to any type of cross-equation restriction, it is based on asymptotic theory, which limits its
usefulness in estimations with small samples (see Enders, 1995; 1996). Likelihood ratio tests are
hence supplemented by multivariate generalisations of the Akaike Information Criterion (AIC) and
Schwartz Bayesian Criterion (SBC).

35
The diagonal elements of βMM and βFF are zero.

112

The equations in 2 are used to test for causality between variables; namely, a standard F-test is used
to check if all the coefficients on the lagged values of a variable included in an equation are jointly
significantly different from zero. A further test is the block causality test, which determines whether
lags of one variable Granger-cause any other variable in the system. Implementing this test involves
restricting all lags of the variable in all the equations of the other variables to be equal to zero.
To further study the interrelationships among the variables in the system, we use impulse response
functions, which capture the dynamic responses of the variables contained in (Xt) to the set of
structural shocks (υMt and υFt). These are obtained by inverting the VAR, yielding its Vector Moving
Average (VMA) representation, such that the variables included in the VAR are expressed in terms of
the current and past values of the n types of shocks:

∞
X t = X + ∑ θ iυ t −i (3)
i =0

By plotting the coefficients of θjk(i) against i we can, for instance, trace the response of the
[ ]
where X = M t , F t is the unconditional mean of X t , n is the total number of variables
in the system and θ i is an ( nxn) matrix with elements θ jk (i) representing impulse response
functions. υ t is an n variate white noise innovation process of X t , such that if t ≠ s, υ t
and υ s are uncorrelated.
macroeconomic variables Mt (REER, inflation and money supply) to flows in aid ( Ft).
However, in order to identify the impulse responses, it is necessary to identify the structural form of
the model from the estimated standard VAR in equation 2. A convenient Vector Moving Average
(VMA) representation is one with orthogonalised innovations since they are uncorrelated both
across time and across equations, hence the economic analysis can be done separately on each
equation. To accomplish orthogonalisation, we use the Choleski decomposition of the covariance
matrix of the VAR model, suggested by Sims (1980); that is, all the elements above the principal
diagonal of the covariance matrix are restricted to zero. This implies an ordering of the variables in
the system such that shocks to each variable contemporaneously affect variables ordered after it but
not before it. With this assumption, the parameters of the structural VAR can be obtained using the
OLS estimates of the standard VAR.
The ordering of the variables is, in this case, done such that variables that we expect to have more
predictive value for other variables are first, for example aid (Ft). However, we take into
consideration the fact that, when there is substantial correlation among innovations in variables, the
decomposition of one-step variance depends strongly on the order of factorisation.

Export Supply Equation
Borrowing from Biggs (2007), we develop an export supply model that assumes supply side
constraints as the major factors affecting exports, including the REER, costs and availability of
working capital. Thus, the long-run equilibrium supply constrains equation is of the following form:

X = α 0 + α 1 ( REER t ) + α 2 ( P xt ) + z t (1)
t

where Xt denotes the logarithm of export supply (volume or value), REERt is the logarithm of the real
effective exchange rate, Pxt is logarithm of export unit price of exports or TOT in case the export

113

equation includes total exports), zt is a disturbance term. It is expected that if the REER decreases,
the supply of exports will fall36, so α1 is expected to be negative. An improvement in the unit price Pxt
(or TOT in the case of total exports) will lead to an increase in the supply of exports, thus α2 is
expected to be positive.
In addition, a number of factors that could affect the supply of exports are included in the short run
specification:
(i) A measure of VAVB. Since exchange rate volatility creates a risky business
environment in which there are uncertainties about future profits and payments for
exporters, the coefficient on VAVB is expected to be negative.
(ii) Rainfall is added to measure the effect of weather on exports. The expected sign on
the rainfall coefficient is positive.
(iii) Credit to the agricultural sector (total credit in the case of total exports) crgt and/or
the lending rate, Lxt. More availability of credit to the sector should promote supply
of exports, while a higher cost of capital would discourage exports.
To establish the long-run equilibrium relationship among the above variables, we assume equation
(1) to be the cointegrating equation, which is estimated using the Johansen methodology. In the
case where no cointegration is established a short run model is estimated.

* 4 i
∆ X t = α ( L ) ∆ X t − 1 + β ( L ) ∆ Ft + ∑ ψ i Z t −1 (2)
i =1

where Zt = Xt –ξ’Ft; Ft = (1, Yt, REERt, v(h)t); Ft* = (Yt, REERt, v(h)t); α(L) and β(L) are lag polynomials and
the vector ξ is the vector of estimated parameters from equation (1). The parameter ψ is the error
correction coefficient.

Estimating Exchange Rate Volatility
As most empirical analysis incorporates a proxy for exchange rate volatility, this could be a source of
inconsistencies of the effect of exchange rate volatility on exports. Some of the previous research
(Kihangire et al, 2005) use a moving average standard deviation of the past monthly exchange rates,
while others (Merton (1980), Klaassen (1999) and Baum et al (1999) use daily spot exchange rates to
compute one month ahead exchange rate volatility. This study uses the GARCH model developed by
Engle, (1982) and Bollerslev (1986) to quantify exchange rate volatility, based on the monthly
nominal exchange rate. Let the ut be the error term of an AR (|1,4|) model of the nominal exchange
rate. Now let the error process be such that:
(3)
u t = vt ht

2
σ v = 1
where
The GARCH model is specified as the conditional variance of ut such that:

2 q 2 p
h t = V ( u t | Ω t −1 ) = α 0 + ∑ α i u t − i + ∑ φ i h + δ ' wt (4)
i =1 i =1 t −1

36
See appendix 8, footnote 8: Methodology for the computation of the NEER and REER for Uganda.

114

Where ht2 is the conditional error variance of ut with respect to the information set Ωt −1 , and wt is a
vector of predetermined variables assumed to influence the conditional error variances in addition
q p
to the past squared errors, while ∑
i =1
α i u t2− i and + ∑
i =1
φ i h t − 1 are the moving average (MA) and

the autoregressive (AR) parts of the model.

115

Appendix 2: Results
Stationarity Test Results – Export Equation
The paper uses Augmented Dickey Fuller (ADF) approach to test whether the series are stationery
either in levels or in first difference. The lag length is determined using both the Akaike information
criterion and Bayesian information criteria. The test results are presented in Appendix Table 2.
Except for maize, rainfall (Rain) unit price of cotton (UnitCot), and value of flowers which are I(0), all
the variables are integrated of order one i.e. I(1), implying that they must be differenced once in
order to make them stationary.
As discussed in Engle and Granger (1987), a linear combination of two or more non-stationary series
can be stationary. If a stationary linear combination exists, then the non-stationary series are said to
be cointegrated. The stationary linear combination is the cointegrating equation and can be
interpreted as a long-run equilibrium relationship among the variables. The presence of a
cointegrating relationship would form the basis for vector error correction estimation. Therefore,
the next step in the analysis is to test for the long-run relationships among the variables in equation
(1) using the methodology developed in Johansen (1991, 1995a). Cointegration was done by
estimating an unrestricted reduced form VAR in levels, including the relevant export variable and its
unit price, and REER. In addition, an unrestricted constant was included. The results of the
cointegration estimations are shown in Appendix Tables 3-7. With the exception of tea, which shows
two cointegrating relationships, no cointegration is found between the REER and the other exports.
Since maize and flowers are stationary, no cointegration tests were performed on them.
Table A1: Augmented Dicker Fuller (ADF): HO: Variable has a unit root
The critical values used in the ADF test are -2.915 and -3.552 at 5% and 1%, respectively.
Variable t- ADF t-ADF 1st diff Order of integration
Log XVALUE -1.591 -5.156** I(1)
Log(coffee) -2.160 -12.2211** I(1)
Log(fish) -2.362 -8.5151** I(1)
Log(REER) -2.638 -7.3085** I(1)
Log(tea) -1.92 -7.9098** I(1)
Log(NER) -1.14 -3.405* I(1)
Log(cotton) -1.728 -10.51** I(1)
Log(maize) -3.598** -8.0920** I(0)
Log (TOT) -2.061 -7.649** I(1)
Log(Rain) -5.641** I(0)
Log (Flowers) -2.104 -5.695** I(1)
Log Credit Agric 0.062 -6.518** I(1)
Log Total Credit -1.471 -5.827** I(1)
Lending rate (LRB) -2.444 -6.479** I(1)
Unit price-coffee -2.061 -6.301** I(1)
Unit price-tea -2.186 -8.329** I(1)
Unit price-fish -1.322 -7.343** I(1)
Unit price-maize -1.78 -9.838** I(1)
Unit price-cotton -3.295* I(0)
Note: **, * means reject the null at the critical values of 1% and 5%, respectively

116

Cointegration Analysis
Table A2: Total Exports (lxvalue)
I(1) cointegration analysis, 1994 (3) to 2006 (4)
eigenvalue loglik for rank
151.0369 0
0.29584 159.8056 1
0.14848 163.8238 2
0.044048 164.9500 3

H0:rank<= Trace test [ Prob]
0 27.826 [0.084]
1 10.289 [0.264]
2 2.2524 [0.133]

Asymptotic p-values based on: Unrestricted constant
Unrestricted variables:
[0] = Constant
Number of lags used in the analysis: 4

beta (scaled on diagonal; cointegrating vectors in columns)
LXVALUE 1.0000 0.32644 0.47996
LTOT -1.7906 1.0000 0.17902
LREER -4.2356 -0.15714 1.0000

alpha
LXVALUE -0.075411 -0.31198 0.044982
LTOT 0.14629 -0.095339 0.046564
LREER 0.028043 -0.026246 -0.043885

long-run matrix, rank 3
LXVALUE LTOT LREER
LXVALUE -0.15566 -0.16890 0.41342
LTOT 0.13752 -0.34895 -0.55809
LREER -0.0015876 -0.084315 -0.15854

Table A3: COFFEE (log Coffee)
40.32116 0
0.22221 46.60366 1
0.11310 49.60424 2
0.015689 49.99957 3

0 19.357 [0.478]
1 6.7918 [0.608]
2 0.79066 [0.374]


117

[0] = Constant

Lcoffee 1.0000 -0.067539 -0.17765
UnitC 4.6758 1.0000 -0.81965
LREER 2.0317 -1.1486 1.0000

alpha
Lcoffee -0.16141 0.80879 0.24018
UnitC 0.069142 2.3636 -0.34950
LREER -0.044999 -0.07305 -0.022161

Lcoffee UnitC LREER
Lcoffee -0.25870 -0.14277 -1.0167
UnitC -0.028406 2.9734 -2.9239
LREER -0.036129 -0.26529 -0.029681

Table A4. Cotton (LCOT)

1.409659 0
0.11347 4.420654 1
0.010282 4.679026 2

0 6.5387 [0.637]
1 0.51674 [0.472]

[0] = Constant

LCott 1.0000 0.09795
LREER -7.8330 1.0000

alpha
LCott -0.37746 -0.50147
LREER 0.016771 -0.013756

LCott LREER
LCott -0.42658 2.4551
LREER 0.015424 -0.14513

118

Table A5. Tea (Ltea)

115.9220 0
0.16420 120.4062 1
0.14668 124.3717 2
0.11634 127.4639 3

0 23.084 [0.250]
1 14.115 [0.079]
2 6.1844 [0.013] *

[0] = Constant

LTea 1.0000 -0.37115 -1.1608
UnitT -1.0483 1.0000 -1.1744
LREER -4.7972 -0.73353 1.0000

alpha
LTea 0.14573 0.23198 0.035442
UnitT 0.074006 -0.14397 0.079700
LREER 0.028021 -0.032033 -0.015214

LTea UnitT LREER
LTea 0.018491 0.037581 -0.83381
UnitT 0.034927 -0.31516 -0.16971
LREER 0.057570 -0.043539 -0.12614

Table A6: Fish (Lfish)
20.64721 0
0.15269 24.78955 1
0.14201 28.61851 2
0.010179 28.87429 3

0 16.454 [0.353]
1 8.1695 [0.225]
2 0.51156 [0.542]

119

Asymptotic p-values based on: No constant

LFish 1.0000 -0.62673 0.56788
LREER -3.6734 1.0000 -0.76285
UnitF 2.8315 0.30042 1.0000

alpha
LFish -0.045982 0.27897 0.00088841
LREER 0.0012243 0.011249 0.0010753
UnitF -0.061682 -0.29033 0.0036741

LFish LREER UnitF
LFish -0.22032 0.44720 -0.045502
LREER -0.0052154 0.0059318 0.0079214
UnitF 0.12236 -0.066555 -0.25820

Error Correction Model
An error correction export determination model is estimated using the saved residuals from the
equilibrium relationships in period (t-1), as the error correction term, and using Ordinary Least
Squares (OLS), focusing on only the equations where the exports of interest are the endogenous
variables. The REER, unit prices/TOT, Rainfall (Rain), credit and/or lending rate and VAVB and
relevant error correction terms are the explanatory terms. All variables except for error correction
terms include four or three lags in the general equation. Following the general to specific modelling,
the results of the reduced estimation results are indicated in Tables A7-A13.

Total Exports
Error! Reference source not found. indicates that the main factors affecting total exports in Uganda
are TOT and volatility of the VAVB. The coefficient on the TOT variable is positive and significant,
indicating that an improvement in TOT would encourage exports. Specifically a 1% improvement in
TOT would lead to an increase in exports receipts of 0.6%. Though the coefficient on VAVB is
significant, it is positive which is not consistent with the fact that exchange volatility would create a
risky business environment that would discourage exports. We do not establish a significant
relationship between the REER and total exports. However, this was not the case for all the
individual export commodities:
Table A7: Total Exports (LXVALUE)
Modelling DLXVALUE by OLS
The estimation sample is: 1994 (3) to 2006 (4)
Coefficient Std.Error t-value t-prob Part.R^2
DLTOT 0.559823 0.2139 2.62 0.012 0.1296
LRain -0.215072 0.04559 -4.72 0.000 0.3261
LRain_1 0.208712 0.04559 4.58 0.000 0.3130
VAVB_1 194.420 81.33 2.39 0.021 0.1105

sigma 0.161524 RSS 1.20014132

120

log-likelihood 22.2927 DW 2.28
no. of observations 50 no. of parameters 4
mean (DLXVALUE) 0.027015 var (DLXVALUE) 0.0420771

AR 1-4 test: F (4,42) = 1.1590 [0.3425]
ARCH 1-4 test: F (4,38) = 0.52995 [0.7144]
Normality test: Chi^2(2) = 1.5303 [0.4653]
hetero test: F(8,37) = 0.65075 [0.7302]
hetero-X test: F(14,31) = 0.87453 [0.5912]
RESET test: F(1,45) = 0.11007 [0.7416]

Coffee Exports
Table A8 indicates that exports of coffee are mainly affected by rainfall and the cost of capital. In
particular, a 1% increase in rainfall leads to a 0.8% increase in coffee exports. An increase in the
lending rate of 1% reduces coffee exports by 0.6%.
Table A8: Coffee
Modelling DLcoffee by OLS
Constant -6.93144 1.609 -4.31 0.000 0.2876
LRain_1 0.417726 0.1407 2.97 0.005 0.1608
LRain_3 0.418414 0.1430 2.93 0.005 0.1568
DLRB_3 -0.595015 0.2814 -2.11 0.040 0.0886

sigma 0.296556 RSS 4.04548817
R^2 0.424851 F(3,46) = 11.33 [0.000]**
log-likelihood -8.08641 DW 2.2
mean (DLcoffee) -0.0281997 var (DLcoffee) 0.140676

AR 1-4 test: F (4,42) = 0.77612 [0.5470]
ARCH 1-4 test: F (4,38) = 0.16267 [0.9559]
Normality test: Chi^2(2) = 15.880 [0.0004]**
hetero test: F(6,39) = 1.8019 [0.1241]
hetero-X test: F (9,36) = 1.1798 [0.3370]
RESET test: F (1,45) = 0.64517 [0.4261]

Tea Exports
Table A9 indicates that tea exports mainly adjust to deviations from long run equilibrium (VLTea_1).
This implies that any shock to tea export production that drifts it away from its long run path would
not last for a long period of time.
Table A9: Tea
Modelling DLTea by OLS
VLTea_1 -0.760893 0.1952 -3.90 0.000 0.2405

121

sigma 0.217802 RSS 2.2770107
mean (DLTea) -0.00923472 var (DLTea) 0.0610996

AR 1-4 test: F (4,44) = 0.25156 [0.9072]
ARCH 1-4 test: F(4,40) = 0.18587 [0.9444]
hetero test: F(2,45) = 0.14963 [0.8615]
hetero-X test: F(2,45) = 0.14963 [0.8615]
RESET test: F(1,47) = 0.81867 [0.3702]

Cotton
Table A10 indicates that cotton exports are affected by the REER, though the effect is very small. In
particular, a 1% increase in the REER (depreciation) is associated with an increase of 0.04% in
exports. The exports of cotton are also negatively affected by their past values.
Table A10: Cotton
Modelling DLCott by OLS
DLCott_1 -0.868083 0.1040 -8.35 0.000 0.6077
DLCott_2 -0.898980 0.1003 -8.96 0.000 0.6409
DLCott_3 -0.696220 0.1039 -6.70 0.000 0.4994
LREER_2 0.0407374 0.0418 0.975 0.335 0.0207

sigma 1.34115 RSS 80.9407045
mean(DLCott) 0.0909413 var(DLCott) 5.34064

AR 1-4 test: F(4,41) = 2.2239 [0.0831]
ARCH 1-4 test: F(4,37) = 1.2305 [0.3147]
hetero test: F(8,36) = 2.2513 [0.0460]*
hetero-X test: F(14,30) = 1.2923 [0.2685]
RESET test: F(1,44) = 7.1652 [0.0104]*

Fish Exports
Table A11 indicates that a depreciation of the exchange rate DL(REER) leads to a 2.7% increase in
fish exports (DLFish). This finding would support the view that an appreciation of the REER would
hurt export competitiveness. It would also have negative implications for poverty reduction, given
that fish is a rapidly growing export (constituting one of the largest merchandise export item) and
being the main source of income for some 266,000 households, equivalent to around 1.2 million
people or 4% of the population. This finding supports the earlier findings by Atingi-Ego and Ssebudde
(2000). Finally, fish export is negatively related to its past values.

122

Table A11: Fish
Modelling DLFish by OLS
DLFish_1 -0.246832 0.1297 -1.90 0.063 0.0715
DLFish_2 -0.313168 0.1291 -2.43 0.019 0.1112
DLREER_1 2.69754 1.115 2.42 0.019 0.1107

sigma 0.362687 RSS 6.18247887
mean(DLFish) 0.0271058 var(DLFish) 0.163697

AR 1-4 test: F (4, 43) = 1.3705 [0.2600]
ARCH 1-4 test: F (4, 39) = 0.18360 [0.9455]
hetero test: F (6, 40) = 1.3145 [0.2732]
hetero-X test: F (9, 37) = 1.0233 [0.4401]
RESET test: F (1, 46) = 0.22424 [0.6381]

Maize Exports
Table A12 indicates that the main factors affecting maize export are rainfall and its past values. The
coefficients on rainfall and maize past values are positive and statistically significant. Specifically, a
1% increase in rainfall would help increase maize export supply by 0.5%.
Table A12: Maize
Modelling LMaize by OLS
LMaize_1 0.519855 0.1147 4.53 0.000 0.2997
LRain_3 0.527776 0.1270 4.15 0.000 0.2645

sigma 0.775469 RSS 28.8648974
mean(LMaize) 9.06644 var(LMaize) 0.897302

AR 1-4 test: F (4, 44) = 1.7419 [0.1579]
ARCH 1-4 test: F (4, 40) = 0.81784 [0.5214]
hetero test: F (4, 43) = 0.34030 [0.8492]
hetero-X test: F (5, 42) = 1.133 [0.3554]
RESET test: F (1, 47) = 0.12753 [0.7226]

123

Flower Exports37
Table A13 indicates that the main factors affecting flower exports are the real effective exchange
rate (DLREER) and cost of capital. Specifically, a 1% increase in the REER would help increase supply
of flower export by 3%. An increase in the cost of capital by 1% would lead to a fall in flower exports
by 1.5%. Rainfall and time (technological change) also affected the export of flowers. More rainfall
discouraged the supply of flowers. In particular, a 1% increase in rainfall reduced supply of flower
exports by 0.3%. Flower exports were also positively affected by developments that occurred over
the review period as reflected by a positive and statistically significant coefficient on the time trend
variable.
Table A13: Flowers
Modelling LFLOWT by OLS
Constant 8.87539 1.202 7.38 0.000 0.6021
DLREER 2.89405 0.8196 3.53 0.001 0.2572
DLRB_1 -0.880685 0.3507 -2.51 0.017 0.1491
DLRB_3 -0.664239 0.3687 -1.80 0.080 0.0827
LRain_3 -0.336656 0.1425 -2.36 0.024 0.1343
Trend 0.0525796 0.0034 15.6 0.000 0.8715

sigma 0.253952 RSS 2.32170273
R^2 0.895761 F(5,36) = 61.87 [0.000]**
mean(LFLOWT) 7.8897 var(LFLOWT) 0.530309

AR 1-3 test: F(3,33) = 0.87931 [0.4618]
ARCH 1-3 test: F(3,30) = 0.90780 [0.4489]
hetero test: F(10,25) = 0.97479 [0.4885]
hetero-X test: F(20,15) = 0.72240 [0.7550]
RESET test: F(1,35) = 3.9411 [0.0550]

Identifying the Effects of Aid
First, we carry out causality tests between the aid (LTFA) and the macroeconomic variables: money
supply, CPI and REER. We estimate two sets of VAR models using equation 2. The first model
includes four variables: aid (LTFA), broad money (BM3), headline CPI, and REER. The second model
includes: aid (LTFA), broad money (BM3), underlying CPI, and REER. Both models include a constant
and seasonal dummy. Our objective is to identify the effects of aid on prices and REER. The aid
variable is placed first in the ordering based on the idea that it would not respond instantly to
contemporaneous movements in other variables. All variables are included in the VAR in log form.
Money supply is included in the system of equations because aid presumably affects prices and
hence REER via changes in money. All variables in the VAR are included in levels, for the reason that

37
The general equation Included the DLREER, Lrain, VAVB, Trend, and LRB.

124

even if differencing would be appropriate, it would yield no asymptotic efficiency gain in an
autoregression. In addition, information would be lost by differencing data since, for instance, co-
integrating relationships among variables would not be captured in the VAR (see for instance Sims,
1980 and Doan, 2000).
Each entry in Table A15 represents a joint significance tests in a VAR, including the aid variable LTFA,
broad money, (LBM3) headline CPI (LCPIH), and REER (LREER). The entries in the first column of
Table A15 suggest an orderly pattern in which aid (LTFA) helps predict (or granger causes) money
supply (Row 3, Col. 3) at a 1% level of significance and helps predict the REER at 10% level. It does
not help to predict itself and CPI. The entries in row 4 suggest that money helps predict (or granger
causes) itself at a 1% level of significance (Row 4, Col. 3) and helps predict the REER at 5% level of
significance (Row 4, Col. 5). It does not help predict aid and prices. The entries in row 5 suggest that
prices help predict the REER but at only 10% level of significance. They do not help to predict the
other two variables (aid and money) in the model. That aid granger causes money but does not
granger cause inflation could be due to fact that due to sterilisation measures of liquidity by the
monetary authorities, injections of liquidity into the economy following government expenditure
would not be inflationary, though they would lead to short run increases in money supply. This is
further reflected by the fact that money supply does not granger cause prices. However, both aid
and money are associated with changes in the REER implying a possible aid induced effect on the
REER, although conclusions cannot be established at this level until we examine the impulse
response functions.
The last row of Error! Reference source not found. indicates marginal significance levels for the
hypothesis that all lags of each of the variables can be excluded from the model. It is shown that all
variables are important in predicting at least one of the other variables in the system. Finally, the
first block of the statistics below Error! Reference source not found. shows single equation statistics
and the second the system tests. None of the tests is significant at the 1% level, except the normality
statistic for DLBM3, which may need reconsideration when a model of the system has been
constructed and evaluated.
Table A15: Identifying the Effects of Aid on REER and Headline Inflation

A. Joint significance tests in VAR including Aid, Broad money 3, CPI-headline, and REER
Dependent Variable

p-values for LTFA LBM3 LCPIH LREER
LTFA 0.282 0.001 0.912 0.098
LBM3 0.603 0.000 0.781 0.015
LCPIH 0.079 0.285 0.000 0.075
LREER 0.336 0.143 0.091 0.000
Marginal significance levels for the
hypothesis that all lags of a variable
can be excluded from the model. 0.000 0.000 0.000 0.000
1. Low probability values in A indicate that at conventional significance levels, the row variable Granger causes
the column variable.
2. Estimates are based on vector autoregressions with 15 monthly lags of each variable.

LTFA : Portmanteau(12): 2.74021
LBM3 : Portmanteau(12): 4.52335
LCPIH : Portmanteau(12): 5.88843

125

LREER : Portmanteau(12): 8.62995
LTFA : AR 1-7 test: F(7,74) = 0.56950 [0.7784]
LBM3 : AR 1-7 test: F(7,74) = 0.49629 [0.8344]
LCPIH : AR 1-7 test: F(7,74) = 1.4099 [0.2142]
LREER : AR 1-7 test: F(7,74) = 0.98327 [0.4500]
LTFA : Normality test: Chi^2(2) = 3.0551 [0.2171]
LBM3 : Normality test: Chi^2(2) = 48.892 [0.0000]**
LCPIH : Normality test: Chi^2(2 = 1.8351 [0.3995]
LREER : Normality test: Chi^2(2) = 1.1715 [0.5567]
LTFA : ARCH 1-7 test: F(7,67) = 0.39424 [0.9025]
LBM3 : ARCH 1-7 test: F(7,67) = 0.46060 [0.8595]
LCPIH : ARCH 1-7 test: F(7,67) = 0.38242 [0.9095]
LREER : ARCH 1-7 test: F(7,67) = 2.2541 [0.0403]*
LTFA : hetero test: Chi^2(120) = 130.29 [0.2454]
LBM3 : hetero test: Chi^2(120) = 138.13 [0.1233]
LCPIH : hetero test: Chi^2(120) = 115.18 [0.6072]
LREER : hetero test: Chi^2(120) = 100.32 [0.9038]
Vector Portmanteau(12): 112.867
Vector AR 1-7 test: F(112,201) = 1.3233 [0.0434]*
Vector Normality test: Chi^2(8) = 55.375 [0.0000]**
Vector hetero test: Chi^2(1200) = 1231.4 [0.2579]

The results obtained are weaker when the underlying CPI is used in the VAR estimation. As Table A16
indicates, aid (LTFA) helps predict (or granger causes) money supply (Row 3, Col. 3) at a 1% level of
significance, but does not help predict the REER, itself and underlying CPI. Money helps predict (or
granger causes) itself (Row 4, Col. 3) and helps predict the REER (Row 4, Col. 5). It does not help
predict aid and prices. Prices do not help predict the REER. They help predict money supply at 5%
level in addition to predicting prices.
Table A16: Identifying the Effects of Aid on REER and Underlying Inflation

A. Joint significance tests in VAR including Aid, Broad money 3, CPI-underlying, and REER
Dependent Variable

p-values for LTFA LBM3 LCPIU LREER
LTFA 0.544 0.000 0.924 0.107
LBM3 0.074 0.000 0.569 0.017
LCPIU 0.049 0.004 0.000 0.147
LREER 0.528 0.241 0.337 0.000
Marginal significance levels for the
hypothesis that all lags of a variable
can be excluded from the model. 0.000 0.000 0.000 0.000
1. Low probability values in A indicate that at conventional significance levels, the row variable Granger causes
the column variable.
2. Estimates are based on vector autoregressions with 15 monthly lags of each variable.

LTFA : Portmanteau(12): 3.28894
LCPIU : Portmanteau(12): 3.28314
LBM3 : Portmanteau(12): 5.87938
LREER : Portmanteau(12): 4.95308
LTFA : AR 1-7 test: F(7,74) = 1.0020 [0.4369]
LCPIU : AR 1-7 test: F(7,74) = 0.54170 [0.8001]
LBM3 : AR 1-7 test: F(7,74) = 0.99715 [0.4403]
LREER : AR 1-7 test: F(7,74) = 0.91015 [0.5035]

126

LTFA : Normality test: Chi^2(2) = 3.6327 [0.1626]
LCPIU : Normality test: Chi^2(2) = 3.6782 [0.1590]
LBM3 : Normality test: Chi^2(2) = 22.376 [0.0000]**
LREER : Normality test: Chi^2(2) = 3.5875 [0.1663]
LTFA : ARCH 1-7 test: F(7,67) = 0.30667 [0.9485]
LCPIU : ARCH 1-7 test: F(7,67) = 0.30782 [0.9480]
LBM3 : ARCH 1-7 test: F(7,67) = 0.57774 [0.7716]
LREER : ARCH 1-7 test: F(7,67) = 0.81315 [0.5796]
LTFA : hetero test: Chi^2(120) = 132.90 [0.1985]
LCPIU : hetero test: Chi^2(120) = 112.24 [0.6805]
LBM3 : hetero test: Chi^2(120) = 135.17 [0.1628]
LREER : hetero test: Chi^2(120) = 105.73 [0.8204]
Vector Portmanteau(12): 100.852
Vector AR 1-7 test: F(112,201) = 0.98182 [0.5374]
Vector Normality test: Chi^2(8) = 35.476 [0.0000]**
Vector hetero test: Chi^2(1200)= 1218.7 [0.3473]

127

Appendix 3: Computation of the REER
Nominal Exchange Rate
This is the nominal exchange rate of the Uganda shilling against the currency of the trading partner
(e.g. US$/Uganda Shilling would for example mean that 1US$ = 1,500 Uganda Shilling, but not the
reverse which is 1 Uganda Shilling = 0.0006.7US$)38. It is obtained from the International Financial
Statistics (IFS) with the exception of the Uganda Shilling to the USA Dollar, which is the monthly
average of the mid-rate. The official mid-rate is the average of the weighted bid and ask rates in the
inter-bank foreign exchange market. Commercial banks’ transaction rates are assigned weights
according to the size of the bank. The weighted averages for all the banks are thereafter summed up
to get a single value, for the weighted buying or selling rates.
The Nominal Effective Exchange Rate
The NEER is an index measure of the local currency against the currencies of Uganda’s trading
partners. It is an index of weighted averages of bilateral exchange rates of the Ugandan shilling in
terms of foreign currencies. The weights are based on trade shares reflecting the relative importance
of each currency in the effective exchange rate basket.
Methodology
k
α
NEER = Π
i = 1
e α
i
i
≡ α
e1 1 X e 2 2
X ..….. X eαk
k

Where;
k = number of major trading partners;
e = the exchange rate of the Uganda Shilling against the trading partner i currency;
αi = The total trade (imports plus exports) weight of country i with Uganda.
k
and ∑α
i =1
i =1

Trade Weights
The weights are derived from direction of trade statistics based on the value of bilateral trade
(imports plus exports) with the trading partner.
Trade weight αi = ti/t
Where;
ti = total volume of trade with country i.
t = total imports and exports of the economy.

38
A decrease in the value of Uganda shillings used to purchase one (1) unit of US Dollar represents an
appreciation of the Uganda shillings against the USA Dollar and vice versa. This therefore means that an
increase in the REER represents a depreciation, while a decrease represents an appreciation.

128

The Real Effective Exchange Rate
The REER of a country is the nominal effective exchange rate adjusted for price differentials between
the domestic and the foreign countries it trades with. Its importance stems from the fact that it can
be used as an indicator of external competitiveness in the foreign trade of a country.
Using the Purchasing Power Parity definition, the REER is defined in the long run as the nominal
effective exchange rate (e) that is adjusted by the ratio of the foreign price level (Pf) to the domestic
price level (P); In Uganda’s case, the core CPI is used to proxy for domestic prices. Mathematically, it
can be shown as
p f
REER = e
p
From this definition, the decline in the REER can be interpreted as the real appreciation of the
exchange rate; the reverse is true in the case of an increase.
Foreign Prices
The foreign prices as used in the REER computation are indices of CPI, or wholesale price indices
(where available) of Uganda’s trading partners, weighted by the trade shares
k
Pf = Π ρ iα i
≡ ρ 1α 1 X ρ α
2
2
X ..... X ρ α
k
k

i =1

Where
k= number of trading partners;
ρ = the Price index (Wholesale or Consumer) of the Country i; and
k
αi = the export trade weight of country i and ∑α i =1
i =1

129

References
ADAM C.S. AND D.L. BEVAN (2002). Uganda: Aid, Public Expenditures, and Dutch Disease. Department
of Economics, University of Oxford, Manor Road, Oxford OX1 3UL.
ARIZE A.C., (1995). “The Effects of Exchange Rate Volatility on US Exports: An Empirical Investigation.”
Southern Economic Journal, Vol. 62, No. 1. pp. 34-43. Southern Economic Association.
ATINGI-EGO M., (2005), “Budget Support, Aid Dependency, and Dutch Disease: The Case of Uganda”,
Paper presented to The World Bank’s Practitioners’ Forum on Budget Support, May 5-6, 2005
Cape Town.
ATINGI-EGO M., AND SSEBUDDE R., (2000). “Uganda’s Equilibrium Real Exchange Rate and its
Implications for Non-traditional Export Performance”, Bank of Uganda Staff Papers, Vol.2, No.
1, 2000, Bank of Uganda, Kampala, Uganda.
BACHMANN, M.O. AND BOOYSEN, F.L.R. (2002). HIV/AIDS and Poverty: Evidence from a Household
Impact Study conducted in the Free State Province, South Africa. Paper presented at DPRU
Conference, 22 – 24 October 2002, Johannesburg, South Africa.
BACHMANN, M.O. AND BOOYSEN, F.L.R. (2004). Relationships between HIV/AIDS, Income and
Expenditure over time in deprived South African households. AIDS CARE. 16(7).
BAILEY, M J., GEORGE S.T., AND M. ULAN. (1987). “The Impact of Exchange Rate Volatility on Export
Growth: Some Theoretical Considerations and Empirical Results”. Journal of Policy Modelling
pp.225-243.
BARNETT, T. AND CLEMENT, C. (2005). HIV/AIDS Impact: so where have we got to and where next?
Development Studies Institute, London School of Economics, London, UK, Progress Report.
BARNETT, T., AND WHITESIDE, A. (2002). AIDS in the Twenty-First Century: Disease and Globalization.
Palgrave Macmillan, Basingstoke, UK.
BAUM C.F, C MUSTAFA AND OSKAN N., (1999). “Non-linear Effects of Exchange Rate Volatility on the
Volume of Bilateral Exports” (PUBLISHER?)
BIDPA (2000). Macroeconomic Impacts of the HIV/AIDS Epidemic in Botswana (Gaborone: BIDPA).
BIGGS T., (2007). “Assessing Export Supply Constraints: Methodology, Data, Measurement”, Revised
June 2007.
BLOOM, E., CANNING, D. AND JAMISON, D.T. (2004). “Health, Wealth and Welfare”, in Health and
Development: Why Investing in Health is Critical for Achieving Economic Development Goals.
International Monetary Fund, Washington, DC, USA.
BULÍR, A. AND A. J. HAMANN (2003). “Aid Volatility: An Empirical Assessment,” IMF StaffPapers, 50(1),
pp: 64-89.
BUREAU FOR ECONOMIC RESEARCH/SABCOHA (2005). The Impact of HIV/AIDS on Selected Business
Sectors in South Africa (Stellenbosch: BER).
COLLIER, P. AND D. DOLLAR (2002). “Aid Allocation and Poverty Reduction”, European Economic Review,
Vol. 46, pp. 1475-500.
COLLIER, P. AND DEHN (2001). “Aid, shocks and Growth”, World Bank Policy Research Working Paper
No. 2688, The World Bank, Washington DC.

130

DEGOL HAILU (2007). “Scaling-up HIV/AIDS Financing and the Role of Macroeconomic Policies in
Kenya”. Conference Paper No 4 (June). Prepared for the Global Conference on
Macroeconomic Policies to Reverse the HIV/AIDS Epidemic, Brasilia, 20-21 November 2006.
UNDP/International Poverty Centre. Brasilia (March). (available at: www.undp-
povertycentre.org).
DEININGER K., AND OKIDI J., (2001). “Rural Households: Incomes, Productivity and Non-farm
Enterprises”, in Uganda’s Recovery: The Role of Farms, Firms, and Government, ed. by Ritva
Reinikka and Paul Collier, World Bank, pp.123-175
DFID (2004). “Macroeconomic Issues for Scaling-up Aid Flows”. Macroeconomic Note No.2 (October).
DOAN, T., (2000). RATS Version 5 User’s Manual. Estima: Evanston.
DOROODIAN K. (1999). “Does Exchange Rate Volatility Deter International Trade in Developing
Countries?” Journal of Asian Economics, 10(3), pp.465-474(10), Elsevier Science.
DTIS (June, 2006). Uganda Diagnostic Trade Integrated Study, report prepared under the Integrated
Framework for Trade-Related Technical Assistance to Least Developed Countries.
http://www.integratedframework.org/countries/uganda.htm
ELBADAWI, I. (1999). “External Aid: Help or Hindrance to Export Orientation in Africa?” Journal of
African Economies, 8(4), pp. 578-616.
EMENYONU, N., MAIER, M., ANDIA-BIRARO, I., KAIDA, A., HOGG R., AND BANGSBERG. D. (2008). ART, Fertility
Desire and Recent Fertility History of HIV+ Ugandan Women (April).
ENDERS, W., (1995). Applied Econometric Time Series. John Wiley and Sons, Inc.: New York,
Chichester, Brisbane, Toronto and Singapore.
ENDERS, W., (1996). Rats Handbook for Econometric Time Series. John Wiley and Sons, Inc.: New York,
Chichester, Brisbane, Toronto and Singapore.
ENGLE R.F. AND GRANGER C.W.J., (1987). “Cointegration and Error Correction. Representation,
estimation and testing”. Econometrica 55, pp. 251-276.
ENGLE, (1982). “Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of
United Kingdom Inflation”, Econometrica 50, pp. 987-1007.
FRIEDMAN, M. (1953). The case for flexible exchange rates. Essays in positive economics. Chicago.
University of Chicago Press, p.175-2003.
GREENER, R. (2004). “The Impact of HIV/AIDS on Poverty and Inequality, in HAACKER, M. (ed.) The
Macroeconomics of HIV/AIDS (Washington DC: IMF).
GREENER, R., K. JEFFERIS, & H. SIPHAMBE (2000). “Impact of HIV/AIDS on Poverty and Inequality in
Botswana”. South African Journal of Economics.
GUPTA, S., R. POWELL & Y. YANG (2005). The Macroeconomic Challenges of Scaling up Aid to Africa. IMF
Working Paper WP/05/179, International Monetary Fund, Washington DC.
GUPTA, S., R. POWELL & Y. YANG (2006). Macroeconomic Challenges of Scaling Up Aid to Africa – A
Checklist for Practitioners. Washington DC: IMF.
HANSEN, H., AND F. TAP, (2002). Aid Effectiveness Disputed, Journal of International Development, Vol.
12, pp. 375-98.

131

INTERNATIONAL LABOUR ORGANIZATION (2004). HIV/AIDS and Work: Global Estimates, Impact and
Response, Programme on HIV/AIDS and the World of Work, International Labour organization,
Geneva, Switzerland.
JAYNE, T.S. (2004). Measuring Impacts of HIV/AIDS on African Rural Economies. Informal presentation
at SARPN, HSRC, Johannesburg, 13 July 2004.
JEFFERIS, K, S. BAINE, J BYAMUGISHA, & J NANNYONJO (2007). Assessing the Macroeconomic Impact of
HIV/AIDS in Uganda - Phase 1 Report: Literature Review: the Macroeconomic Impact of
HIV/AIDS (Kampala: UNDP & MoFPED).
JEFFERIS, K., A. KINGHORN, H. SIPHAMBE & J. THURLOW (2006). The Economic Impact of HIV/AIDS in
Botswana (Gaborone: UNDP/National AIDS Co-ordinating Agency).
JOHANSEN, S. (1991). “Statistical Analysis of Cointegration Vectors in Gausian Vector Autogressive
Models”. Econometrica, 59(6): pp.1551-1580.
JOHANSEN, S. (1995a). Likelihood Based Inference in Cointegrated Vectors Autogressive Models,
Oxford University Press, Oxford.
JOHNSON, G.H. (1969). “The Case for Flexible Exchange Rates,” Federal Reserve Bank of St. Louis.
Monthly Review, June 1969. 51(b), pp. 12-24.
KIHANGIRE D, AND D.J POTTS AND CAMEROON S. (2007). “Has Exchange Rate Variability Affected Tropical
Non-traditional Exports: Evidence from Uganda’s Flowers Exports (1994-2001)”. The Bank of
Uganda Staff Papers Journal Vol. (1), No.2, 2007 pp.49-63. Bank of Uganda, Kampala Uganda.
KLAASEN F. (1999). “Why is it so Difficult to Find an Effect of Exchange Rate Risks on Trade? A VAR
Approach”, Review of Economics and Statistics, 68, pp.311-315.
KRUGMAN, P. (1989). “Exchange Rate Instability.” The Lionel Robbins Lectures. Cambridge. The
Massachusetts Institute of Technology (MIT) Press.
LAKE, S & MWIJUKA, B. (2006). Sector-based Assessment of AIDS Spending in Uganda – Final Report
(European Commission, November 2006)
MERTON, R. (1980). “On Estimating the Expected Return to the Market. An Explanatory
Investigation”. Journal of Financial Economics, 8, pp.323-361.
MINISTRY OF FINANCE, PLANNING AND ECONOMIC DEVELOPMENT (MoFPED) (2004). Poverty Eradication
Action Plan (PEAP) 2004/5 -2007/8.
MINISTRY OF FINANCE, PLANNING AND ECONOMIC DEVELOPMENT (MoFPED), Background to the Budget.
Several issues (2004-2007) June. MFPED, Kampala. (available at: www.finance.go.ug).
MINISTRY OF FINANCE, PLANNING AND ECONOMIC DEVELOPMENT AND BANK OF UGANDA (2005). Export Sector
Performance and Exchange Rate Appreciation. Mimeo.
MINISTRY OF HEALTH (2003). Anti-retroviral Treatment Policy for Uganda (Kampala: MoH).
MINISTRY OF HEALTH (2006). Uganda HIV/AIDS Sero-Behavioural Survey 2004-2005 (Kampala: MoH).
MINISTRY OF HEALTH (2007). Compiled Statistics on HIV/AIDS Prevalence and ART (Kampala: MoH).
MUNDELL, R. A. (2000). Currency Areas, Exchange Rate Systems and International Monetary Reform,
Journal of Applied Economics. III, pp.217-256.

132

NAMPANYA-SERPELL, N. (2000). “Social and Economic Risk Factors for HIV/AIDS Affected Families in
Zambia: some Research Findings”. Paper presented at the AIDS and Economics Symposium, 7-8
July 2000, Durban, South Africa.
NANNYONJO, J. AND J. O. APAA, (2007). Effects of Exchange Rate on Export: Case Study of Uganda.
Unpublished.
NKUSU M (2004a). “Financing Uganda’s Poverty Reduction Strategy: Is Aid Causing More Pain than
Gain?” IMF Working Paper WP/04/179 (Washington DC: IMF).
NKUSU M. (2004b). “Aid and the Dutch Disease in Low-income Countries: Informed Diagnoses for
Prudent Prognoses” IMF Working Paper WP/04/49 (Washington DC: IMF).
OKWARE, S., OPIO, J M & WAIBALE, P.( 2001). “Fighting HIV/AIDS: Is Success Possible” Bull. World Health
Organ, Vol. 79, No.12.
ONI, S.A., OBI, C.L., OKORIE, A., THABEDE, D., AND JORDAN, A. (2002). “Economic Impact of HIV/AIDS on
Rural Households, Smallholder Agricultural Production and Orphans in the Limpopo Province
of South Africa”. The South African Journal of Economics, 70 (7).
PALLAGE, S. AND M. A. ROBE (2001). “Foreign Aid and the Business Cycle,” Review of International
Economics, 9, pp: 636-668.
RAJAN, R. & A. SUBRAMANIAN (2005). “What Undermines Aid’s Impact on Growth,” IMF Working Paper
05/126, Washington DC: IMF.
ROBALINO A., JENKINS, C. AND EL MAROUFI, K. (2002). “Risks and Macro-Economic Impacts of HIV/AIDS in
the Middle East and North Africa: Why Waiting to Intervene can be Costly”. Policy Research
Working Paper No 2874. World Bank, Washington DC, USA.
ROSEN, S, J. VINCENT, W. MACLEOD, M. FOX, D. THEA, & J. SIMON (2004). “The Cost of HIV/AIDS to
Businesses in Southern Africa”, AIDS, 18:2.
SALINAS, G. & M. HAACKER (2006). “HIV/AIDS: The Impact on Poverty and Inequality”, Working Paper
WP/06/126 (Washington DC: IMF).
SEKKAT, K. AND A. VAROUDAKIS (2000). “Exchange Rate Management and Manufactured Exports in Sub-
Saharan Africa”, Journal of Development Economics, 61(1), pp: 237-253.
SERIEUX, J. (2007). Managing the Exchange Rate Consequences of an MDG-related Scale-up in
HIV/AIDS Financing. Paper prepared for the Global Conference on Macroeconomic Policies to
Reverse the HIV/AIDS Epidemic, Brasilia, 20-21 November 2006. UNDP/International Poverty
Centre. Brasilia (March). (available at: www.undp-povertycentre.org).
SIMS, C.A., (1980). “Macroeconomics and Reality.” Econometrica, Vol.48: pp.1-48.
STEINBERG, M, JOHNSON, S., SCHIEROUT, G., & NDEGWA, D. (2002) Hitting Home. How Households Cope
with the Impact of the HIV/AIDS Epidemic (Washington DC: Kaiser Foundation).
STOVER, J. (2005). AIDS Impact Model Manual (Policy Project).
STOVER, J., AND S. KIRMEYER (2005). Spectrum Model Manual (Policy Project).
THE COMPETE PROJECT AND ALICE GOWA (2001). The Path Forward in Uganda’s Coffee Sector. Prepared
on behalf of the European Commission for the Presidential Conference on Export
Competitiveness.

133

TSIKATA, T, M. (1999). “Aid Effectiveness: A Survey of the Recent Empirical Literature”, AERC Research
Paper No. 110, Economic Research Consortium, Nairobi.
UBOS (2007). Uganda National Household Survey 2005/6 (Kampala: UBOS).
UGANDA AIDS COMMISSION (2001). Twenty Years of HIV/AIDS in the World: Evolution of the Epidemic
and Response in Uganda. Kampala.
UGANDA AIDS COMMISSION (2007). National HIV/AIDS Strategic Plan 2007/08-2011/12. (Kampala:
UAC).
UGANDA BUREAU OF STATISTICS (UBOS) (2003). Uganda National Household Survey 2002/2003. Report
on the Labour Force Survey. (Kampala: UBOS).
UNAIDS (2002). 2002 Report on the Global AIDS Pandemic, Joint United Nations Programme on
HIV/AIDS, Geneva. Switzerland.
UNAIDS (2004). 2004 Report on the Global AIDS Pandemic, Joint United Nations Programme on
HIV/AIDS, Geneva. Switzerland.
UNDP (2007). Uganda Human Development Report 2007 (Kampala: UNDP).
USAID SOUTH AFRICA/BUREAU FOR ECONOMIC RESEARCH (2006). The Macroeconomic Impact of HIV/AIDS
Under Alternative Intervention Scenarios (with Specific Reference to ART) on the South African
Economy (Johannesburg: USAID).
WEIDLE, P. et al (2002). “Assessment of a Pilot Anti-retroviral Drug Therapy Programme in Uganda:
Patients’ Response, Survival and Drug Resistance” . The Lancet, Vol. 360, Issue 9326, pp.34-40.
WYSS, K., HUTTON, G., AND N’DIEKHOR, Y. (2004). “Costs Attributable to AIDS at Household Level in
Chad”, AIDS CARE , 16 (7).

134

Phase II – Selected Studies

More Related Content

What's hot (17)

Similar to Phase II – Selected Studies (20)

More from econsultbw (20)

Phase II – Selected Studies