ECONOMIC VALUATION OF ANTIRETROVIRAL

DRUGS IN ETHIOPIA: APPLICATION OF CVM

BY
MARTHA GETACHEW

A Thesis Submitted to the School of Graduate Studies of Addis

Ababa University in Partial Fulfilment of the Requirement for
the Degree of Master of Science in Economics (Human
R es o u rce)

ADDI S AB AB A
J UNE , 2 0 0 3
 ACKNOWLEDGMENT

I would like to extend my deepest gratitude to my advisor Dr. Abdulhamid

Bedri Kello for his invaluable guidance.

I am forever indebted to my family, especially Getiye and Wudit. Thanks for

the love and encouragement you gave me without any reservation.

I would also like to acknowledge Kassahun (Dawn of Hope) and Mekonen

Alemu (Mekdim Ethiopia) for the excellent cooperation during the survey. I
am also indebted to Dr Agonafer Tekalegn of CRDA for all the
encouragement and Dr. Steven Forsythe for sending me his dissertation related
to contingent valuation method on HIV/AIDS programmes.

I would like to acknowledge Christian Relief and Development Association

(CRDA) and African Economic Research Consortium (AERC) for providing
me with financial support to complete this thesis. My heartfelt gratitude in
particular goes to the latter for awarding me a scholarship throughout my
study.

My appreciation extends to all who offered me moral support. I am especially

grateful to my colleagues Solomon Tesfay, Samuel Faye, Bisrat Egigu and
Tadele Ferede.
TABLE OF CONTENTS
Page
Acknowledgment
List of Tables
List of Figures
List of Abbreviations
Abstract
Chapter 1: INTRODUCTION………………...………………………….1

1.1 Background……………………………...…………………………..1
1.2 Statement of the problem. ………….……………………………….2
1.3 Objectives of the Study…………………………………………...…6
1.4 Hypothesis of the Study……………………………………………..7

Chapter 2: ARV: AN OVERVIEW…………………………………...….9

Chapter 3: LITERATURE REVIEW…...………………………...……21

3.1 Methods of Economic Valuation…………………...………...……21

3.2 The Contingent Valuation Method………………………………....26
3.2.1 Theoretical Foundation of WTP Approach…………….….....26
3.2.2 Methodological Basis……………………………………..….29
a) Bias Issues………………………………………..………30
b) Elicitation Techniques………………………...………… 35
3.3 Empirical Literature Review………………………………………..37

Chapter 4: METHODOLOGY………………………………..…...……43

4.1 Data Source and Type……………………………………….….….43

4.2 Sampling Procedure……...………………………………………...44
4.3 Questionnaire Design……………………………...…………….....45
4.4 Methods of Data Collection……………………………...………...47
4.5 Method of Data Analysis: An Econometric Approach…...……..…47
4.6 Model Specification………………………………………..…...….55
Chapter 5: EMPIRICAL FINDINGS AND ANALYSIS………..…….59

5.1 Descriptive Analysis…………………...…………………………..59

5.1.1.Soico-Demographic Characteristics……..………………....59
5.1.2 KAP…………………………………………….………..…61
5.1.3 Willingness to pay…………………………………….……63
5.1.4 Economic Status……………………………………………66
5.1.5 Health related indicators…………………………………....67
5.1.6 General Attitude……………………………………....…….70
5.1.7 Validity Test…………………………………………..……71
5.1.8 Starting Point Bias Test………………………………….....73
5.1.9 Free Riding Test………………….……………………..….73
5.2 Regression Analysis…...……………………………………...……74
5.2.1 Parametric Estimation ……………….…………………...…..74
5.2.2 Data Property ………….…………………….…………….....77
5.2.3 Semiparametric Estimation……………………………...……79
5.2.4 Summary of Estimation Results……...………...…...…..……84
5.2 Demand for ARV drugs……………………………..…………...…86

Chapter 6: CONCLUSION AND POLICY IMPLICATION…....……90

6.1 Conclusion……………………………………………………..…..91
6.2 Limitation of the Study………………………………………….…93
6.3 Policy Implication…………………………………………...……..93

References………………………………………………………….……...96
Appendix 1: Results of Chapter five……………….………………….....101
Appendix 2: Results of Data property……………………………….…...107
Appendix 3: Questionnaire………………………………………....……110
 LIST OF TABLES

Page

Table 1.1 Current and Future HIV/AIDS infected population and adult
prevalence rate, 2002 and 2010……………………………...….4
Table 2.1 Classification and Functions of ARVs……………………….....11
Table 2.2 Too early and too late prescriptions of ARVs………………..…13
Table 4.1 Expected signs and relationships of variables………………..…56
Table 5.1 Socio-demographic characteristics of respondents………...….101
Table 5.2 Age and family size of respondents…………………………...101
Table 5.3 Knowledge and practice of respondents……………………....102
Table 5.4 Attitude of respondents towards ARVs………………….……102
Table 5.5 Reasons for not paying…………………………………….…..102
Table 5.6 Willingness to pay amounts……………………...………..…..103
Table 5.7 Summary statistics of willingness to pay……………………...103
Table 5.8 Willingness to pay and initial bid price…………………….....103
Table 5.9 Willingness to pay and occupational status of respondents...…103
Table 5.10 Test for free riding…………………………………...……....103
Table 5.11 Correlation matrix of income and WTP………………...……104
Table 5.12 Result of the Tobit estimation…………………………………75
Table 5.13 Result of the CLAD estimation………………………….…….82
Table 5.14 Results of the 25th and 75th quantile regressions………….….106
Table 5.15 Estimated total WTP and demand……………...…………..….86
 LIST OF FIGURES
Page
Figure 1.1 Essential enzymes and ARV drugs……………...……………..12
Figure 3.1 Economic Valuation………………………………………...…24
Figure 4.1 Bidding game with Birr 350 to Birr 150 starting points……….46
Figure 5.1 Estimated demand curve for ARV drugs………………...….....87
 LIST OF ABBREVIATION

AIDS Acquired Immune Deficiency Syndrome

ANC Antenatal Care
ART Antiretroviral Therapy
ARV Antiretroviral drugs
CVM Contingent Valuation Method
DOH Dawn of Hope
HAART Highly Active Antiretroviral Therapy
HIV Human Immunodeficiency Virus
MOH Ministry of Health
NGOs Non-governmental organizations
OI Opportunistic infection
PLWHA People living with HIV/AIDS
PMTCT Prevention of Mother to Child Transmission
TB Tuberculosis
UNAIDS Joint United Nations Programme on AIDS
UNWECP United Nations Workplace Education and Care Project
WHO World Health Organization
WTA Willingness to Accept
WTP Willingness to Pay
 ABSTRACT

In this study, the willingness to pay of people living with HIV/AIDS for anti-

retroviral drugs is assessed. A contingent valuation survey was carried out on

440 adults living with HIV/AIDS. The research result indicates that the mean

willingness to pay for the drugs is Birr 163.6 and the median is Birr 100, per

month. A multivariate statistical analysis of the determinants of individual’s

willingness to pay reveals that individuals with larger family size, with others

paying for their treatment cost other than themselves and with spouse

deceased are less willing to pay for the drugs. Individuals with higher income

are more willing to pay. Older respondents are also willing to pay more.

Among the explanatory variables, income was found to be highly significant

in influencing WTP by the different estimation techniques employed in the

research implying low income levels of respondents prevent them from

acquiring the drugs, if the drugs be marketed. This calls for the participation of

many stakeholders in the acquisition and distribution of anti-retroviral drugs if

equityisdesired.
Key words: Contingent valuation method, Willingness to pay, antiretroviral

drugs, CLAD

CHAPTER ONE

INTRODUCTION
1.1 BACKGROUND

The HIV/AIDS pandemic continues to be an alarming problem for Ethiopia where 2.6

million, including about 250,000 children are HIV positive and an estimated 400,000

Ethiopians have already developed AIDS. The number of PLWHA is growing rapidly in

the country in general and in Addis Ababa in particular. The incidence of voluntary HIV

testing and counselling as well as early entry into the continuum of care directed at

PLWHA is very low. These circumstances are aggravated in Addis Ababa by a general lack

of awareness of the confidential services that are available to HIV/AIDS clients (UNAIDS,

2002). Of the total residents of Addis Ababa, 17% live with HIV/AIDS. In other words, one

in six adults (15-49 years old) is already infected with the virus. Frighteningly, a vast

majority of those already infected are unaware of their status.

It is estimated that by 2004, of the total death in Addis Ababa, 60% will be an outcome of

AIDS- this in other words means out of five people that die, three would be due to AIDS
1
(MOH, 2000). While the epidemic results death on people in all age range, the effect it has

on those between 15 to 49 years old is overwhelming. By 2009, it is estimated that 80

people in this age group will die every day because of AIDS. The consequence of such a

death toll will definitely have an adverse effect on the socio-economic development of the

country.

There is no cure for HIV/AIDS, but in developed countries, AIDS is being turned into a

manageable and treatable illness instead of death sentence with the use of life prolonging

HIV/AIDS drugs. Data from the United States show that the combination therapies known

as highly active anti-retroviral therapies (HAART) reduced AIDS –related mortality by

75% and AIDS- related morbidity by 73% over three years. Similarly in the Brazilian state

of São Paulo, AIDS-related mortality has declined by 32% since protease inhibitors became

available in 1996 (FHI, 1998).

However, in the African region, the number of HIV infected people who can access and

afford ARV therapy is still very small. Of the 25 million people that are living with the

virus, only 10,000-25,000 people were receiving ARV therapy in the year 2001

(UNAIDS/WHO, 2001). The high price of many AIDS drugs –especially ARV drugs- is

one of the main barriers to their availability in developing countries. Fortunately, the price

of ARV drugs is coming down (discussed in detail in the next chapter) as a direct response

by major drug companies to the UNAIDS and WHO ARV Accelerated Initiatives in several

countries. Similarly, some companies in Brazil, Thailand and India are producing generic

products that are much cheaper and affordable to more people in the developing countries.

2
Hence, it is anticipated that in the next few years there will be a bigger number of people in

Africa who will be accessing ARV drugs.

1.2 STATEMENT OF THE PROBLEM

Ethiopia is ranked as the 168th country in human development, where most of its population

has “low access” to essential drugs. In terms of priority in public expenditure, over a nine

years period (1990-1999) the public expenditure on health slightly increased from 0.9% to

1.2% of GDP. On the other hand, the total expenditure on military expenditure increased

from 8.5% to 9.4% of GDP over the same period. The Gini1 index, which measures

inequality over the entire distribution of income or consumption, indicates that it is 55.1%

for the country. With regard to human poverty index, the country is ranked as 87th out of 88

developing countries (UNDP, 2002).

With regard to the HIV profile of the country, Ethiopia is the 16th highest in HIV/AIDS

prevalence and the third largest number of people living with the virus next to South Africa

and India. Official reports estimate that 2.2 to 2.6 million people are currently infected

with HIV/AIDS where around two million are adults and 200 to 250 thousand children

(MOH, 2002 and UNAIDS, 2002). While the number of new AIDS cases reported to the

Ministry of health in 2001 are 15,202 this figure was estimated to be as high as 219,400.

The peak ages for AIDS cases are 25 to 29 for both males and females. This implies that

given the average incubation period between the time of infection and emergence of full

3
blown AIDS is about eight years, the mean age at which people become infected is 15 to 24

years.

Of the total infected population, about 91% of infections occur among the economically

active age group, 15-49 years. This indisputably has a huge social and economic impact on

the country as a whole.

For 2001, the estimated adult HIV prevalence rate was 6.6% where the highest prevalence

was observed among the age group 15-24. Compared to rural prevalence rate, which is

3.7% the urban prevalence rate is double this figure, 13.7%.

In 1989, the life expectancy of Ethiopians was estimated at 45 years, expected to be 53

years by 2001 and 59 years by 2014. However, due to the increase in HIV-related deaths,

instead of 53 years in 2001 the life expectancy declined to 46years and 50 instead of 59 by

2014.

Ethiopia is identified to be hit by the Next Wave of HIV/AIDS among four other countries -

Nigeria, Russia, India and China (NIC, 2002). By 2010, it is estimated that the five

countries, which comprise over 40% of the world’s population, will have the largest

number of HIV/AIDS case on earth if nothing is done presently. The following table shows

the estimates of current and future HIV/AIDS-infected population and adult prevalence rate

in the Next-Wave countries for 2002 and 2010.

1
The Gini coefficient will be equal to 0 when the distribution is completely egalitarian while if the society’s
total income accrues to only one person/household unit, leaving the rest with no income at all the coefficient
will be equal to 1, or 100%.
4
 Table 1.1. Current and future HIV/AIDS-infected population and adult prevalence rate,

2002 and 2010

2002 2010

Country # of infected # of infected Adult # of infected Adult

Population population Prevalence (Expert Prevalence
(government (Expert rate (%) estimates rate (%)
data in millions) estimates in millions)
in millions)
China 3.5 4-6 6-10 10-15 18-26
Ethiopia 2.2* 3-5 10-18 7-10** 19-27
Russia 0.18 1-2 1.3-2.5 5-8 6-11
India 4.0 5-8 0.9-1.4 20-25 3-4
Nigeria 0.8 1-2 0.14-0.27 10-15 1.3-2
Source: NIC, 2002
* MOH, 2002 **MOH estimates this figure to be 2.9
From the above table it can be seen that Ethiopia’s adult prevalence in the future (19%-

27%) will be the highest among the five countries. Because of the high current rate of adult

prevalence, widespread poverty, low educational levels, and the government’s limited

capacity to respond more actively, 7-10 million Ethiopians will probably be infected by

2010 (NIC, 2002).

The development of ARVs helps to improve the survival and quality of life of many people

living with HIV/AIDS. Despite the remarkable success of the drugs, however, their prices

are very expensive. According to UNAIDS (2002), the projected annual expenditure

requirements for HIV/AIDS care and support by 2005, excluding the costs of infrastructure

required for delivery of effective HIV treatment are US$ 4,400 million. Out of this more

than half, 69%, is taken by sub-Saharan Africa. The estimated figure for South and South

East Asia is 15%, East Asia and Pacific 1.8%, Latin America and Caribbean 12 %. The

5
remaining 0.45% and 1.13% go to East Europe and Central Asia, and North Africa and

Middle East, respectively.

In particular, it is estimated that treatment of all people living with HIV/AIDS with

HAART could represent 1% of the GDP in Latin America, 14.6% in South East Asia and

66.9% in the sub-Saharan Africa (Walkowiak, 1999).

It is argued that the provision of a fully subsidized ARV treatment program by the

Ethiopian government is not considered affordable. However, it is believed that it is

affordable to the country to have a fully government financed prevention of mother-to-child

transmission programme, which is estimated to cost 2.28 million USD per year (Policy

framework, 2002).

Clinical trials for vaccine against HIV are underway and have progressed to phase III

human testing in the US, Canada, Netherlands, and Thailand. However, a vaccine is

developed to protect against certain viral subtypes of HIV in the test countries, which

probably will not be effective against the most common subtypes of HIV in countries which

will be hit by the Next Wave of HIV/AIDS-Nigeria, Ethiopia, Russia, India and China

(NIC, 2002). This leaves the countries to focus on anti-AIDS education to bring

behavioural change to prevent new cases and ARV drugs to prolong the lifetime of people

already living with the virus until a highly effective vaccine is developed for these

countries.

The direct result of the inaccessibility of ARV drugs in health structures in Ethiopia is the

development of ARV distribution channels outside the health care system. Currently,

6
although ARV drugs are not officially available in the country, some patients are

nonetheless receiving ARV treatment (UNAIDS, 2002). The drugs are traded in the

“parallel market” for prices from Birr 1,000 to Birr 5,000 (US$115-$575) per month.

1.3 OBJECTIVES OF THE STUDY

The Ethiopian government is considering provision of the drugs for free only in the case of

prevention of mother-to-child transmission (PMTCT) of the virus. The provision of ARV,

when the drugs are available in the market, will be based on user fees for other individuals

living with the virus. With respect to this the Ethiopian government seems to strike a

middle ground between abandoning all other drugs and buy only ARVs and distribute the

drugs for free to all PLWHA, and leaving this totally to the private sector.

The objectives of the study are:

To estimate the willingness to pay of people living with HIV/AIDS for anti-retroviral

drugs

To identify the determinants of willingness to pay of people living with HIV/AIDS

(PLWHA) for the drugs

To estimate the demand function of respondents for antiretroviral drugs

HYPOTHESIS OF THE PAPER

7
The hypothesis of the paper on the value of life is based on the assumption stated by

Johansson (1995). Individuals prefer to stay alive even if this virtually means with a zero

income i.e.,

V (p, 0, z>0) > 0

V (p, y, z=0) =0 ∀y>0

Where: V = the indirect utility function

p = fixed, strictly positive prices

y = income

z = individual’s health status which is “normalized” so as to range between zero

and some finite positive level, which is often arbitrarily set equal to one, where 0 is the

worst possible health status and 1 is the best possible health status.

In other words, V (p, 0, z1) > V (p, y, z0) where z1 means being alive, z0 means being dead,

and y > 0. Thus, it is not possible to define a willingness to pay such that the inequality is

turned into equality i.e., even if the individual pays his/her entire income the inequality

remains.

The specific study hypotheses to be tested are:

• There is no demand for antiretroviral drugs in Ethiopia

• Major determinants of willingness to pay for antiretroviral drugs are income,

employment status, years of living with the virus and treatment cost of opportunistic

8
 infections whereas most socio-demographic characteristics of individuals are not

significant factors for determining the willingness to pay for ARV drugs.

CHAPTER TWO

ARV DRUGS: AN OVERVIEW

HIV is a retrovirus acquired through contact with body fluids, primarily through sexual

intercourse and Utero-from infected mothers to newborn children or contact with infected

blood. The virus survives by replicating inside CD42 blood cells, which normally protect

the body against infection. Currently two families of this virus are known- HIV-1 and HIV-

2 where the latter was discovered three years later. These exhibit extensive genetic

diversity, although both result in AIDS. The transmission rate of HIV-2 appears to be lower

compared with HIV-1 and the rate of progression to disease may also be slower.

3
There are four phases of HIV infection . Soon after entering the body, HIV starts to

replicate and destroy CD4 cells. Generally a decrease in CD4 count is correlated with HIV

disease progression, which can be identified by running a blood test. This first phase is

2
CD4 test, cluster differentiation test, could be used to measure how much damage has been done by HIV to a
person’s immune system.
3
Mainly extracted from Sande and Moelling, (2000) . Guide to HIV/AIDS Therapy, 9th ed.

9
called the acute HIV syndrome stage; in this stage the virus causes flu-like symptoms

and/or a rash, which disappear after a few days.

A few weeks after the first stage, the viral load (level of virus in the body) levels gradually

decrease and CD4 cell counts rise again. This phase of the disease is called the

asymptomatic stage. Patients sometimes remain in this stage of infection for several

months or even years. During this time, the virus continues to replicate and attack CD4

cells, although the harmful effects of HIV on the immune system do not necessarily cause

symptoms. As the virus continues to replicate in the second stage, the viral load eventually

increases and the CD4 cell count decreases once more.

In the third phase, the symptomatic stage, patients may begin to develop recurring

symptoms of enlarged lymph nodes, weight loss, fever, diarrhoea, rashes, sores, or oral

infections.

Full-blown AIDS is the last stage of HIV infection in which people consequently suffer

from certain conditions, including opportunistic infections- infections that develop due to

lowered immunity. People with AIDS may die from these opportunistic infections rather

than from the virus itself.

Although HIV infection is incurable and there is no vaccine that has been proven to prevent

the virus, it does not mean that there is no way of prolonging life for carriers of the virus.

Anti-retroviral drugs boost the immune system and help it to fight HIV infections. This can

mean that instead of becoming sicker, a person with AIDS can regain good health. If the

drugs are taken properly they work by suppressing the HIV virus so that it drops to a very

low level in the body. When this happens, the immune system gradually recovers and

10
regains its ability to fight infections, which is reflected by a rise in the CD4 count. In

healthy adults there are between 700 and 1500 CD4 cells per cubic millimetre of blood. A

count below 500 is considered a sign of a depressed immune system; below 200 heightens

the risk of opportunistic infections; below 100 signifies severe immune damage (Panos,

2001). Studies show that dramatic reductions in the viral load can be achieved by using

ARV drugs.

There are currently three types of anti-retroviral drugs: Nucleoside Reverse Transcriptase

Inhibitors, Non-Nucleoside Reverse Transcriptase Inhibitors, and Protease Inhibitors.

The ideal treatment strategy for people living with HIV/AIDS is triple combination, which

involves the use of one protease inhibitor and two reverse transcriptase inhibitors

(Unicef/UNAIDS, 2001).

Studies show that using triple combination therapy reduces not only viral load by 99% but

also the rate of mortality due to HIV/AIDS by as much as half. Among those in triple

combination therapy, 65 to 81 percent had reduced their level of virus to undetectable

levels after six months of treatment (Forsythe, 1998).

Table 2.1 Classification and function of ARVs

Type of ARV drugs Classification Function

Zidovudine (ZDV, also Nucleoside RTI* - Prevention of MTCT** of HIV
Known as AZT) -Also used in combination therapy
Nevirapine (NVP) Non Nucleoside RTI -Prevention of MTCT
-Used in combination therapy
Abacavir Nucleside RTI Used in combination therapy
to suppress replication of HIV

11
 Delavirdine mesylate Non-Nucleoside RTI ”
Didenosine (ddI) Nucleoside RTI ”
Efavirenz Non-Nucleoside RTI ”
Indinavir Protease Inhibitor ”
Lamivudine (3TC) Nucleoside RTI ”
Nelfinavir Protease Inhibitor ”
Ritonavir Protease Inhibitor ”
Saquinavir Protease Inhibitor ”
Stavudine (d4T) Nucleoside RTI ”
Zalcitabine (ddC) Nucleoside RTI ”
Source: Sande and Moelling, 2000.
*RTI = reverse transcriptease inhibitors
**MTCT = mother-to-child transmission
Therapeutic antiretroviral drugs target essential enzymes of the virus to supress HIV/AIDS

as shown in the following diagram:

Fig. 2.1 Essential enzymes and ARV drugs

Viral RNA is copied into Viral DNA integrates Viral proteins are cut
DNA by virus specific into the by the virus-specific
enzyme chromosomes of the PROTEASE for
REVERSE target cell. assembly of the new
TRANSCRIPTASSE INTEGRASE particle and its
maturation

Marketed Drugs Marketed Drugs Marketed Drugs

AZT None Squinavir

Portease Inhibitors
Nucleoside RTI

DDI Indinavir
D4T Nelfinavir
DDC Amprenavir
3TC
12
 Abacavir

Nevirapine
Non-Nucleoside RTI Delavirdine
Efavirez

Source: Unger et al. (2000)

Improvement of a patient’s quality of life is determined by the starting time of taking the

ARV drugs. Prescribing the drugs both too early and too late in a patient’s illness has its

own adverse effects. This is especially true for patients with asymptomatic HIV infection.

Table 2.2. Too early and too late prescription of ARVs

Prescription Benefits Risks

-Avoid negative effects on quality -Possible risk of irreversible
of life (i.e., inconvenience immune system depletion
from taking the drugs) -Possible greater difficulty in
Delayed - Avoid drug-related adverse effects suppressing viral replication
- Delay in development of drug -Possible increased risk of HIV
resistance transmission.
- Preserve maximum number of
available and future drug options
when HIV disease risk is highest
-Control of viral replication easier to - Drug-related reduction of
achieve and maintain quality of life
Early -Delay or prevention of immune - Greater cumulative drug
system compromise related adverse effects
-Lower risk of resistance with complete -Earlier development of drug

13
 viral suppression resistance if viral suppression is
-Possible decreased risk of HIV sub-optimal
transmission. -Limitation of future ARV
treatment options.
Source: HIV 101, 2001

Therefore, if the drugs are prescribed too early, the drugs’ side effects may be noxious,

while the health of the person living with the virus may deteriorate if prescription is too

late. Most guidelines focus on CD4 counts because it is possible to begin therapy by

determining the infection level. Even if the US and the British ARV guidelines mandate

commencement of ARV therapy if a patient’s CD4 count is below 350 and recommend

treatment if the count is in the range of 350 to 500, clinical trial and observational data

indicate that since the risk of opportunistic diseases increases markedly when the CD4 cell

count declines to <200 cells/mm3, patients with this level of CD4 cell count or clinically

defined AIDS should be offered the therapy (HIV101, 2001). This is also supported by

Gebregiorgis (2001) who argues that least developed countries should stick to a CD4 count

of <200, as this is the only proven standard for treatment. However, in resource-poor

settings CD4 and viral load counts are unlikely to be available on all patients. When to start

ARVs is then determined by clinical symptoms such as oral candidiasis, weight loss greater

than 10%, and chronic diarrhoea (Cock, 2001).

The assumed positive impacts of ARVs are reduction in the need for inpatient hospital

visits; the probability of an increase in the productivity of individuals as workers extend

their time within the workforce; the reduction of the number of new infections. This

assumes that a reduction in viral load leads to reduced infectivity and also that there is a

14
decline in risky behaviour, including an increase in the practice of voluntary counselling

and testing. However, recent studies indicate that for a certain percentage of all people

living with AIDS, ARVs are ineffective (Forsythe, 1998). It may also induce people to

become careless and spread the disease if HIV infection is perceived to be a treatable

disease. Moreover, the costs of ARVs may far exceed any economic benefit.

It is generally accepted that the most effective form of treatment is a combination of three

ARVs, including one protease inhibitor. However, because the protease inhibitors are new

and relatively expensive, many developing countries don’t view providing access to this

full complement drugs as an affordable option (Forsythe, 1998). Floyd and Gilks (1997)

identified that triple combination therapies’ cost range from US$ 662 to $993 per month,

double combination therapies range in costs from US$ 403 to $773 per month, while the

one therapy, Nevirapine, costs US$272 per month. At the beginning of year 2000 the price

of combination ARV drugs to treat one patient for one year was typically between US$

10,000 to $12,000 almost everywhere (UNAIDS, 2001).

Generally, medical drugs are produced by two kinds of companies:

Pharmaceutical companies primarily undertake R&D and hold patents in their new

discoveries, which only they have the right to manufacture and sell. Most of these

pharmaceuticals also known as originators or the research-based industry are based in

the North and dominated by few large multinational companies.

15
 Generics Manufactures, on the other hand, primarily produce medicines based on

the formulae of the patented companies. They may do some additional research to

produce their own versions. These are based both in the North and the developing world

(Panos, 2002). Generic or non-patented drugs can be much cheaper than patented ones

because in the case of imitation the cost stream is much smaller than investing in

innovation.

The primary goal of both pharmaceutical and generic companies is to make profit. For

pharmaceuticals, profit comes from the sales of patented drugs. Even if pharmaceutical

companies recognize that ARVs are very expensive to be affordable for developing

countries, they are also concerned that if they offer discounts there might be arbitrage

opportunities (Forsythe, 1998). For generics, profit generally comes form selling drugs that

are non-patented, no longer patented or patented elsewhere- at prices lower than those of

their competitors. Generic ARVs are being produced in countries that ignore patents, such

as Brazil, India and Thailand. However, these countries are not free from pressure from the

giant companies.

To ensure steady future demand for the drugs, the pharmaceutical companies ask

developing countries to sign extended contracts for the purchase of ARVs. Recently,

however, prices offered to low-and-middle income countries fell dramatically.

Two international attempts have been made to increase access to the drugs. The first effort,

Accelerating Access Initiative, was launched in May 2000 by the joint UN programme on

HIV/AIDS (Panos, 2002). By the end of 2000, prices of US$500 to US$800 were being

16
negotiated by low- and-middle income countries for therapies based on patented and

generic drugs. By December 2001, certain generic combinations were on offer for as low as

US$350 per person per year (Unicef/UNAIDS, 2001). The second international attempt to

increase access to drugs was the Global Fund for HIV/AIDS, Tuberculosis and Malaria.

This was launched in October 2002, which announced that generic drugs could be

purchased with its grants (Panos, 2002).

Recently, five leading pharmaceutical companies (Boehringer Ingelheim, Bristol-Myers

Squibb, Glaxo Wellcome, Merck and co., and HoffmanLa Roche) have dropped their third

world prices for components of HAART, which for those who can afford has turned to be

good news.

ARV T RE AT M E NT S AT M A CR O L E VE L

Analysing ARV treatments at macro level could involve - affordability, cost effectiveness,

potential of cost saving and access and equity.

Affordability: this is assessed by obtaining information on the number of people eligible

for ARV treatment, ARV treatment cost per person and total resources of the country. A

WHO/UNAIDS (1998) estimation of the total cost required to have an ARV treatment

programme to treat the whole adult HIV/AIDS infected people in Ethiopia reveals that this

will cost 1.35 billion USD per year. If only people with actual AIDS cases are to be treated,
17
then 200 million USD per year is required. This does not include the total cost required for

capacity building – the human and physical resources required for ARV treatment. These

are:

• Sufficient number of counsellors

• Counsellors trained in drug adherence counselling

• Sufficient number of mid wives for MTCT interventions

• Laboratory services capable of offering the sophisticated tests required for monitoring

ARV treatments

• Extra clinic space for counselling and testing services

• Adequate number of physicians trained in the usage of ARVs

• Patient follow-up to ensure compliance with drugs and/or manage side effects

• Strengthened health education concerning safe feeding practices for mothers receiving

treatment to prevent MTCT.

• Secure buildings for drug storage

• Drug resistance monitoring capacity etc (WHO/UNAIDS, 1998).

According to one survey in Addis Ababa carried out by UNWECP in 2001, the majority of

hospitals and higher clinics in the city deliver fragmented services to HIV/AIDS

clients/patients. Not all hospitals and higher clinics have laboratories for HIV screening or

confirmatory testing. Furthermore, the majority of health facilities do not use supportive

laboratory services to conduct staging of the disease or to monitor the progress of ARV

treatment in their patients. There are only five licensed diagnostic facilities that perform

limited types of HIV tests. Of these facilities only two conduct CD4 cell count and viral

load tests (UNAIDS, 2002) .

18
The total cost of capacity building in Ethiopia is estimated to be 52 million USD per year

(Policy Framework, 2002). The Ethiopian government therefore does not consider a fully

subsidized ARV treatment program to be affordable to the country. However, the total cost

of treatment and testing for the prevention of MTCT program is estimated at 2.28 million

USD per year, which is believed to be affordable to the country. The cost of PMTCT

programmes was estimated by using the following figures:

Proportion of women that have access to antenatal care (35%ANC coverage)= 1,140,000

The cost for screening (1.5USD per screening)=1,710,000

Total cost for treatment = 570,000

Total cost for treatment and testing cost 570,000+1,710,000 = 2.28 mill USD

Cost Effectiveness: the decision to use ARV treatment by an individual is based on the

willingness and ability to pay. However, the decision whether ART is likely to represent a

good use of resources managed on behalf of communities is based on cost-effectiveness

analysis. The guidance module on ARV treatments of WHO/ UNAIDS (1998) suggests that

prevention of MTCT is the most cost effective. Triple combination therapy for infected

adults, however, is cost effective in countries with per capita incomes over USD 8,000.

Cost Saving Potential: If ARV treatments prevent HIV- related health problems like TB,

then they will result in some cost savings. However, Floyd and Gilks (1998) note that

therapies would only be cost saving in countries with average per capita incomes of at least

US$ 8,776.

19
 Equity and Access: This is concerned with the distribution of resources for ARV

treatment. Equity addresses the issue of who is likely to benefit from ARV treatment, which

has an important implication for financing policy. The second issue addressed is the

fairness of such an access. WHO/UNAIDS (1998) guidance module states that since ARV

treatments require specialized staff and equipment, urban populations are more likely to

have access to them than rural populations and some marginalized and vulnerable groups.

With respect to financing, different options can be considered including public funding,

national insurance and social security schemes, private insurance, voluntary sector or

charitable funding, local or community funding, out-of-pocket payment /private funding/,

donor financing and development loans. The WHO/UNAIDS guideline states that:

“unless the public sector can afford to provide a substantial portion of the funding required

for ARV therapy, or unless social insurance is wide spread, it is unlikely that provision to

anything but a small minority of patients is feasible” p. 24

THE FDRE POLICY ON ARV DRUGS

The ARV drugs supply and use policy of the Federal Democratic Repulic of Ethiopia was

formulated in 2002, fifteen years after one of anti-retroviral agent (AZT/Zidovudine) was a

common medical practice to prescribe to AIDS patients in other parts of the world.

The objectives of this policy include:

• Reducing transmission of HIV from mother to child.

20
 • Prolonging and improving the quality lives of PLWHA thereby making them

productive and contribute to the well being of their families and development of

their country.

• Reducing accidental HIV infection within health care institutions.

With regard to supply strategies of ARV drugs the policy, among some other more, outlines

the following:

The government shall supply ARV drugs used for prevention of MTCT to the

appropriate health care institutions through coordination of all stakeholders.

ARV drugs shall be exempted from taxation; supplied at a reduced price through

government’s negotiation with manufacturers, importers and distributors; and be

purchased through a system of bulk and generic substitutions.

The private sector will be encouraged to be involved in locally manufacturing,

importing and distributing ARV drugs.

21
 CHAPTER THREE

LITERATURE REVIEW
The first section of this chapter describes some of the methods of economic valuation

including revealed preference method, stated preference method and human capital

approach. This will be followed by a review of the theoretical foundation of

willingness to pay. The last section of the chapter reviews empirical CV studies in

health economics.

3.1 Methods of Economic Valuation

Several practical methods can be used to measure willingness to pay for goods and services

in general and health services in particular. The three most frequently used and/or
22
suggested methods are indirect methods using market data, survey techniques, and the

human capital approach (Johansson, 1995). The first two are the most common ways of

estimating economic values attached to non-marketed goods and services. If a good or

service has positive economic value, then preferences show up through individuals’

willingness to pay (WTP) for the good or service in question. WTP in the market is made

up of two components: the price or what is actually paid and the consumer’s surplus or the

excess of WTP over the price. The latter is then a measure of the net gain from the purchase

of a marketed good. On the other hand, in a pure non-market context, all WTP is

consumer’s surplus because there is no market price. A disservice, or “bad”, has negative

economic value shown up by the WTP to avoid the bad in question, as a willingness to

accept (WTA) compensation to tolerate the disservice (Summary Guide, 2002). For cost

benefit analysis based on the Hicks-Kaldor compensation test, WTP would seem to be the

appropriate measure for gainers from some resource allocation decision, and WTA the

proper measure for losers from that same allocation (Bateman and Turner, 1997).

Market-Based / Revealed Preference/ Methods

These methods are referred to as indirect methods since the preference revelation is indirect

via a market. In these methods market prices are used to evaluate risks. For example, the

economic value of noise nuisance can be reflected in house prices where houses in noisy

areas are likely to be cheaper than comparable houses in quieter but otherwise similar areas

(Summary Guide, 2002). This method is used to study individuals’ averting behaviour

where by they buy themselves a risk reduction for money. WTP can be estimated either by
23
observing the prices that people pay for goods in various markets or by observing

individual expenditures of money or time to obtain goods, or to avoid their loss, and by

inferring WTP amounts (Abelson, 1996). Revealed preference uses direct observation of

actual values for complementary effects. The different approaches used to measure value

impacts are:

Travel cost model, which is used to value recreational assets via the expenditures on

travelling to the site. The idea is to collect information about preferences from people’s

actual behaviour. In other words, people do not buy the commodity (visit an area) unless

they find it worth its price (travel cost). However, this method captures only use values and

does not consider planned future visits (Freeman, 1998).

Direct choice model or random utility model uses choices between alternative options

to reflect the well-being/ utility that accrues from those options.

Averting behaviour estimates expenditures involved to avoid unwanted effects. An

instance is the presence of an air conditioner in a home or in a car to reduce exposure to air

pollution. However, not only the cost of averting behaviour is inherently difficult to

measure but also measuring the cost of averting behaviour is complicated by the fact that

many avoidance activities produce joint products.

Hedonic Pricing measures the effects that show up in labour markets (or property

markets). This is the most common indirect method used to study wage differentials in the

labour market. Given firms’ and workers’ institutional environment, wage differentials can
24
be explained as firms’ different offers of wages depending on what health risks are

involved, and workers’ different preferences for safety. If a firm has undertaken risk

reduction measures, the maximum wage it is willing to offer will be lower than otherwise.

In general, some argue that revealed preference studies, whether based on labour market

data or consumer behaviour, make untested assumptions about individuals’ risk perceptions

(Krupnick et al, 2000).

Stated Preferences Method

This is the modern name for the survey method. Stated preference approaches are based on

constructed markets. That is, under this approach people are asked to state the economic

value they attach to those goods and services under consideration. The stated preference

approach elicits economic values by using two methods: the contingent valuation method

where respondents to a questionnaire are asked for their WTP and choice modelling which

refers to a variety of procedures for inferring WTP from sets of rankings or ratings of

alternative options presented to respondents (Summary Guide, 2002).

Mitchell and Carson (1993) acknowledge Robert K. Davis (1963) to be the first to use

CVM for the estimation of the benefits of outdoor recreation. This method provides the

only way of directly measuring willingness to pay and willingness to accept.

The following figure summarizes the concepts that have been introduced so far.

25
 Fig.3.1 Economic valuation

Marketed goods and Non-Marketed goods

services and services

E Economic Value =Market Stated Preferences from

price + constructed markets Revealed Preferences
consumer’s from complementary
markets
surplus

Contingent Valuation Choice Modelling

Total Economic Value = Total user

value + Total Non-user value

Summary Guide, 2002

From the above figure it is obvious that the objective of economic valuation technique is to

uncover the total economic value of the good in question. The value may accrue to users-

persons who make direct or indirect use of the good and to those who make no direct use of

the good (non-users); an instance for the latter is the WTP for endangered species, even

though the individual may not have seen, or expect to see, the species in question.

Human Capital Approach

Under this view, the value of an individual is equal to the value of his/her contribution to

total production, and assumes that this value can be measured as his/her earnings

26
(Johansson, 1995). This implies that the value of preventing someone’s statistical death or

injury is equal to the gain in the present value of his/her future earnings.

Value = ∑Πt+i Et+i / (1+r)i ∀ i= 1…..T

Where: Πt+i = probability of the individual surviving from age t to age t+i

Et+i = expected earnings of the individual at age t+i

r = discount rate

T = age of retirement from the labour force

According to this approach. The value of preventing the death of an individual who is

presently at age t is the discounted present value of that individual’s earnings over the

remainder of his/her expected life (Freeman, 1998).

The implication of the approach is that not only is the statistical life of retired people has

zero value but also the statistical life of children is likely to be worth less than that of adults

in or near their best period of earnings. The latter is due to the fact that because of

discounting and the time lag before children become productive participants in the

economy, the human capital approach places a much lower value on saving children’s lives

compared with saving the lives of adults in their peak earnings years. People whose value

of production is not reflected by wage payments (like housewives) are also difficult to

handle within the human capital approach framework. Another shortcoming of this method

is it does not take into account the indirect damage due to death and injuries. Furthermore,

this approach is highly criticized for being not consistent with the individualistic foundation

of welfare economics since it does not take people’s own preferences on changes in the

health risks into account.

27
3.2 THE CONTINGENT VALUATION METHOD

Contingent valuation is a method of estimating the value that a person places on a good or

service. Rather than inferring from observed behaviours in regular market places, the

approach asks people to directly report their willingness to pay to obtain a specified good or

service, or willingness to accept to give up a good or service.

3.2.1 Theoretical Foundation of WTP Approach

When an estimation of WTP is carried out for marginal changes in the probability to die

within a given period of time then one is dealing with an ex post scenario while it is ex ante

if the concern is with the WTP for marginal changes of the expected life span (Zweifel,

1997).

Let’s consider a simple model in which individual’s utility is a function of risk of death and

wealth (Y). Survival (φ) is designated by a dummy variable where,

φ= 1 if “life”

0 “death”

Hence, U(φ, Y)

The assumptions made by this model are:

28
 i. U[1,Y] > U[0, Y] i.e., the individual at any level of wealth prefers life to death.

ii. Uy[1, Y] >Uy[0, Y] Where: Uy = ∂U(φ, Y)/ ∂Y

iii. Uy[φ, Y] >0, Uyy [φ, Y] <0 i.e., U (φ,Y) is twice continuously differentiable, strictly

increasing, and strictly concave.

Assumption ii states that the marginal utility of wealth in case of survival is higher than in

the case of death.

U1(Y) = U[1,Y], U0[Y] = U[0, Y]………………….…………………(1)

The above are the two conditional utility functions, which can be written as:

U (φ, Y) = φ U1(Y) + (1-φ) U0(Y)……….………………….………….(2)

If the dummy variable (1-φ) is replaced by Π, the probability of death, then the expected

utility of the individual can be obtained by weighing the utility payoffs obtained in the

states of nature “life” and “death” by their respective probabilities and adding them up.

Hence,

EU(Π, Y) = (1-Π)U1(Y) + Π U0(Y)…………..……………..……… (3)

The total derivative of (3) gives us:

∂EU ∂EU
dEU [Π, Y ] = dY + dΠ = 0 ……………….………………(4)
∂Y ∂Π

{
= (1 − Π )U1' [Y ] + Π U 0' [Y ]}dY + (U 0 [Y ] − U1[Y ])dΠ

29
To get the marginal rate of substitution between changes in risk and wealth, let’s solve with

respect to dY/dΠ,

dY U1 [Y ] − U 0 [Y ]
m[Π, Y] = = > 0 ……………..………………(5)
dΠ dEU = 0 (1 − Π )U1' [Y ] + ΠU 0' [Y ]

The above equation states how much the individual’s wealth must increase in order to

compensate him/her for an infinitesimal increase in the risk of death (or decrease in the

probability of survival). In other words, the monetary value of a marginal reduction of the

death risk corresponds to the individual’s marginal rate of substitution between his/her

wealth and his/her probability of survival. The numerator shows the utility difference

between “life” and “death” at a given level of wealth, which is positive due to assumption

(i) and the denominator shows the expected value of the marginal utility of wealth, which is

positive as well, due to assumption (iii).

If we differentiate dY/dΠ with respect of Π then,

∂m 2
=d Y =
( ) > 0 ………………………..(6)
− (U1[Y ] − U 0 [Y ]) U 0' [Y ] − U1' [Y ]
∂Π dΠ 2
{(1 − Π )U [Y ] + ΠU [Y ]}
'
1
'
0
2

implying the response of the marginal rate of substitution to an increase in the initial level

of the death risk Π. The expression is positive because the first term in parentheses in the

numerator is positive due to assumption (i) and the second is negative due to assumption

(ii). Therefore, the marginal rate of substitution and hence the marginal WTP for

infinitesimally small changes in death risk rises with an increase from its initial level.

30
Similarly, to examine how the marginal rate of substitution, m(Π, Y), changes with a

variation of the initial level of wealth, let’s differentiate (5) with respect to y

∂m(Π , Y ) =
( )
EUy U1' [Y ] − U 0' [Y ] − (U1[Y ] − U 0 [Y ])EUyy
> 0 ………………………(7)
∂Y (EUy )2

Where: EUy = ∂EU = (1 − Π )U1' [Y ] + Π U 0' [Y ] > 0

∂Y

2
EUyy = ∂ EU = (1 − Π )U1'' [Y ] + Π U 0'' [Y ] < 0
∂Y 2

In line with assumptions (i) and (ii), the numerator on the right hand-side of (7) is positive

and so is the denominator. Therefore, the marginal WTP for a reduction of the death risk

increases with wealth.

3.2.2 Methodological Basis

A contingent valuation survey generally provides a direct way of eliciting WTP values by

asking a sample of households what they would be willing to pay for a good or service.

Telephone survey, mail survey, and in-person interviews are the various methods used to

directly query individuals about their WTP. Telephone and in-person interviews allow

respondents to ask questions whereas not only is this not possible in mail survey but also

this method does not provide one the luxury of conducting follow-up questions about WTP

depending on the answers given to previous questions. It is argued that telephone and mail

surveys are less cheap, particularly in developed countries, but produce less quality WTP

data. In-person interviews in general help to produce the highest quality WTP data, though

31
 very expensive since these are labour intensive. In most LDC situations telephone and mail

surveys may not be practical, leaving in-person surveys as the method of choice.

a) Bias Issues

CVM surveys are prone to various types of bias, which occur in the design as well as in the

execution of the survey.

 Hypothetical bias: If the researcher does not create a believable and meaningful set of

questions that will simulate a market for the good/service in question with sufficient

plausibility or if the scenario is not consistent with reality, then the responses people give to

the hypothetical questions are likely to be careless, which do not reflect their taste

preferences. Bateman and Turner (1997), however, argue that WTP formats unlike WTA

formats, do not suffer from hypothetical bias because usually respondents are familiar with

payment rather than compensation scenarios.

 Strategic bias/free riding: If respondents believe payments of their WTP will be collected

from them, and believe also that there is a good chance the good/service will be provided

even if they understate their true WTP amounts, they tend to underbid their WTP - this is

strategic classic free rider. On the other hand, respondents may ‘over pledge’ i.e., they

would be expected to overbid if they believe they will not actually have to pay the amount

they state, yet believe that the stated amount can influence provision of the good/service.

32
 Starting point bias: This type of bias occurs when the respondent’s WTP amount is

influenced by a value introduced by the scenario. Confronted with a Birr figure where s/he

is uncertain about the value of the good/service in question, a respondent may regard the

proposed amount as conveying an approximate value of the good’s/service’s true value and

anchor her/his WTP amount on the proposed amount. Hence such “bidding hint” lead

respondents to take cognitive short-cuts to arrive at a decision rather than thinking seriously

about their true WTP. Therefore, this is an effect such that the final amount at the end of a

bidding game (see next section) is systematically related to the initial bid value.

 Payment Vehicle bias: WTP may depend on choice of payment vehicle where different

payment vehicles may cause strategic behaviour. This bias occurs when the payment

vehicle is either misperceived or is itself valued in a way not intended by the researcher.

Bateman and Turner (1997) suggest that researchers can effectively eliminate such

problems by avoiding controversial payment methods and instead use that vehicle which is

most likely to be used in real life to elicit payment for the good in question.

 Question ordering bias/ sensitivity of sequencing: WTP varies with where in a sequence

of goods a particular good appears. This is potentially troublesome in CV surveys when

values are elicited for several different programmes or different levels of a programme. For

some goods there is a natural sequence to the provision of the good that prescribes the order

of questions, and respondents answering later questions are intended to take their answers

to earlier questions into account. Here the question order may influence the amounts given

33
 for one or more of the goods that are being valued, despite the interviewer’s instructions to

the respondents to disregard previous answers.

 Interviewer and respondent Bias: Respondents may have a tendency to answer with yes in

order to express motivation or just to please an interviewer for a programme instead of

giving true preferences. For instance, if the interviewer is highly educated or attractive the

respondent may feel inhibited about expressing a low WTP bid. Hence in this Yea-saying

bias, elicited WTP are too high. This can be minimized by using mail or telephone rather

than face-to-face method of interview. On the other hand, in Nay-saying bias, respondent

may have the tendency to counter the interviewer if the former dislikes the latter. In this

case too low WTP are elicited.

 Non-response Bias: Respondents may refuse to answer or they may give ludicrously high

WTP or untrue zero WTP (Protest Zeros). This may be because they oppose the payment

vehicle or are ambivalent. If protest zeros are removed from the sample, higher WTP may

result.

 Warm glow effect: Also known as mental account bias or part-whole bias. WTP may not

vary with the size of the programme because respondents do not reveal preferences but

general approval. That is, the individual’s WTP responses fail to distinguish between

specific good which is under analysis and the wider group of goods into which that specific

good falls. The dimensions of a good that are particularly prone to this misperception are its

geographic distribution, its benefit composition, and the package of policies of which it is a

34
 part. Geographical part-whole occurs when a respondent values a good whose spatial

attributes are larger or smaller than the spatial attributes of the researcher’s intended good.

Benefit part-whole bias is related to the fact when a respondent includes a broader or a

narrower range of benefits in valuing a good than intended by the researcher. If a

respondent values a broader or a narrower policy package than the one assumed by the

researcher then it is policy-package part-whole bias.

 Information bias: Respondents may not be familiar with the commodity being valued or

WTP may vary with information provided. Bateman and Turner (1997), however, note that

the important factor is to ensure that such information is seen to be true, constant across the

sample, and not designed to induce bias towards a particular result, polemic and implicit

value judgement being inadmissible.

Other biases that may occur during a CV survey include: Framing effects bias where WTP

depends on how the question is framed. For instance, whether the glass is seen as half full

or half empty (Summary Guide, 2002); Preference imprecision bias occurs when

respondents are unable to cite precise WTP; Reference dependency bias refers to the

situation when WTP varies with the reference point as perceived by the respondent.

Apart from the afore mentioned problems of CVM, Whittington (2002) tried to categorize

the sources of problems in CVM when applied in a developing country. He identified three

main reasons why many CV studies conducted in developing countries are so bad. The first

is poor survey implementation due mainly to poorly trained enumerators and the resulting

enumerator bias. This is the principal-agent problem in which the researcher (principal)

35
typically does not know the enumerators (the agents) before the survey and has no long-

term relationship with them.

The second problem identified by Whittington is poorly crafted contingent valuation

scenarios. This is the same as hypothetical bias, in which case CV researchers themselves

cannot construct hypothetical choices that make economic sense to respondents.

Failure to test for the effect of variations in survey design on the results of contingent

valuation studies is the third problem with CV surveys conducted in developing countries.

For instance in iterative bidding (see next section) researchers may not do split-sample

experiment to test whether respondents’ WTP would be influenced by the magnitude of the

initial price that they received and the sequence of follow-up questions. However, in my

opinion these biases may also be common in developed countries and could not be argued

to be peculiar for developing countries.

In general, although there are several sources of bias in CVM and some particularly

pronounced in developing countries, Abelson (1996) notes that most of them can be dealt

with appropriate survey techniques. Similarly, Bateman and Turner (1997) also highlight

that CVM surveys can be designed to reduce bias problems to an acceptable level.

b) Elicitation Techniques

By using different formats it is possible to elicit WTP from a sample of households.

According to different literatures on CVM, it seems that no elicitation method is exempt

from one or more than one bias discussed in the previous section.
36
 Open-ended format: Individuals may be asked to state their maximum WTP or

minimum WTA amount with no value being suggested to them. For instance, respondents

could be asked “What is the maximum you would be willing to pay for….?”. This

elicitation technique produces continuous bid variables. The method is straightforward and

no anchoring bias occurs because it does not provide cues about what the value of the

change might be. The other advantage of the technique is it is very informative since

maximum WTP can be identified for each respondent. However, the format leads to

unrealistically large bids and generally unreliable responses. This is because it might be

very difficult for respondents to come up with their true maximum WTP ‘out of the blue’.

In addition, free riding behaviour may occur, which will result in a downwardly biased

estimates of WTP.

 Referendum/ Take-it-or-leave-it/close-ended bidding: Under this format

individuals are presented with a single payment that they accept or reject. This method asks

“Are you willing to pay $X?” It is argued that this method is incentive compatible i.e., it is

in the respondent’s strategic interest to accept the bid, if his/her WTP is greater than or

equal to the price asked or reject otherwise so that the respondent is telling the truth. Even

though it minimizes non-response and avoids outliers, there could be some degree of Yea-

saying. Apart from the existence of starting point bias it is criticized for being inefficient in

that less information is available from each respondent i.e., the researcher only knows

whether WTP is above or below a certain amount. Hence, larger samples and stronger

statistical assumptions are required, which make surveys expensive.

37
  Double-bounded dichotomous choice: the question might run as “Would you pay

$X amount?” If the response is yes then the next question is “ and would you pay $X+Y”

or “and would you pay $X-Y” if the response to the first question is no. The actual WTP

then lies between X and Y if s/he accepted to pay $X in the first question but rejected

$X+Y in the second. Though the double bounded dichotomous choice method is more

efficient than single bounded dichotomous choice as more information is elicited about

each respondent’s WTP , all the limitations of the latter procedure still apply. Other cons of

this technique include anchoring and yea- saying biases in addition to the possible loss of

incentive compatibility (truth telling) due the fact that the second question may not be

viewed by respondents as being exogenous to the choice situation.

 Iterative Bidding: under this elicitation format individuals are asked to respond

either to increasing figures until they reach their maximum WTP or to reducing figures if

they are not WTP the initial figure suggested. It is believed that this method facilitates

respondents’ thought process and encourages them to consider their preferences carefully.

Because of the initial bid prices usually anchoring bias exists. Moreover, it may lead to a

large number of outliers and yea saying. Another drawback of the method is it cannot be

used in mail surveys and other self completed questionnaires.

 Payment card elicitation technique: this format is developed by Mitchell and Carson in

1986, as an alternative to bidding game (Mitchell and Carson, 1993). Respondents are

provided with a range of values from which to select. The question would be like “which of

the amounts listed below best describes your maximum WTP?” One advantage of this

38
method over bidding game is it avoids starting point bias. Bateman and Turner (1997),

however, indicate that anchoring of bids within the range given in the card may occur with

most respondents assuming that such a range contains the “correct” valuation (range bias).

In comparison to bidding game, payment card elicitation method reduces the number of

outliers. Though values in the card serve as a benchmark, there may be vulnerability to bias

relating to the range of the numbers used in the card and the location of the benchmarks.

3.3 EMPIRICAL LITERATURE REVIEW

CVM techniques have been widely used for the purpose of estimating environmental

benefits in particular. Recently, however, a lot of studies have tried to assess health

care/services by applying CVM. Johansson (1995) emphasizes that CVM is appropriate for

health studies since biases are minimized and the problems in using the CVM are somewhat

smaller, or at least different, when the method is applied to health care, which is primarily a

private good than when it is applied to large environmental changes or other public projects

mainly characterized by externalities.

Different researchers used CVM to elicit WTP for different programmes. For instance,

Cameron (1988) surveyed the WTP of patients, physicians, and managed care executives

for diagnostic certainty for peptic ulcer disease and gastroesophageal reflux disease; Bishai

and Lang (2000) elicited the WTP for a month reduction in waiting time for cataract

surgery using bidding elicitation method; Pol and Cairns (1999) applied dichotomous

choice question with follow up to examine inter-temporal preferences for non-fatal changes
39
in own health; Tambour and Zethracus (1998) used CVM to estimate the WTP for hormone

replacement therapy; Bonato et al (2001) elicited the WTP for three possible health care

programmes dealing with Alzheimer’s diseases which is an acquired syndrome of decline

memory and at least one other cognitive function most common among aged people.

The review of 42 health care contingent valuation studies by Diener et al (1998) that were

published between 1984-1996 shows the use of CVM in valuing different health care

programmes is increasing. The reviewers classified health care interventions as medical,

surgical, pharmaceuticals and hypothetical- upon which the CVM is applied. Some of the

studies were performed to assist in pricing and demand forecasts for a product and some to

assist with programme evaluation. The latter was specially concerned with the implication

of the introduction of a new health care programme or the removal of an existing

programme. The vast majority of the studies reviewed by Diener et al (1998) have been

compensating variation- WTP designs. The review of the 42 studies indicted that 55% of

contingent valuation studies in health care surveyed persons either currently diseased or

undergoing the treatment intervention- ex post analysis; 36% of the studies obtained

responses from non-users that are at future risk of contracting the disease or in need of the

intervention- ex ante analysis, while 10% of the studies asked respondents from the general

population which may include respondents from the afore mentioned groups. The majority

of the studies not only used target groups but also valued certain outcomes, and valued

goods from a purely private market perspective. It was also found that all of the elicitation

methods described in the above sub-section were employed almost equally by the 42

studies reviewed by Diener et al. (1998).

40
In 1997 and 1998, two studies estimated the WTP for reduced incontinence symptoms. The

first was by Johansson et al (1997) in Sweden where 461 patients with urge or mixed

incontinence were asked to state their WTP for a reduction in incontinence symptoms using

mail survey. The result showed that in terms of magnitude of risk reduction, patients with

incontinence problems were willing to pay more for a 50% than 25% reduction in the

number of micturitions and leakages. The second study conducted in the United States by

O’Conor et al (1998) also found that the mean WTP for a 50% reduction to be much higher

(US$ 245) than for a 25% reduction (US$ 88) in micturitions and leakages.

Another health care study using CVM was done by Krupnick, et al (2000). In this intensive

study, they administered a mortality risk reduction survey on 930 persons aged from 40 to

70 years in Ontario. By treating risk reduction as a private good (i.e., estimating each

respondent’s WTP to reduce his/her own risk of dying), they estimated what older people

would pay for a reduction in their risk of dying in addition to assessing the impact of health

status on WTP. They reported that age has no effect on WTP until roughly age 70 and

above and the health status with the exception of having cancer, has no effect. However, the

researchers found that being mentally healthy raises WTP substantially. In this study the

elicitation technique used was follow-up dichotomous choice to obtain WTP for risk

reductions of a given magnitude, occurring at a specified time. In one sub sample,

respondents were first asked if they were WTP for a product that, when used and paid for

over the next ten years, will reduce baseline risks by 5 in 1000 over the next 10 years

period (5 in 10,000 annually). In the second WTP question, risks were reduced by 1 in 1000

41
(1 in 10,000 annually). This question was used to serve two purposes- on the one hand, it’s

possible to test whether respondents were willing to pay anything today for a future risk

reduction and on the other hand, to test for internal consistency of responses because WTP

today for a future risk change should be less than WTP today for an immediate risk change.

Their finding suggests that responses were consistent with economic theory, and that WTP

was sensitive to the magnitude of the risk reduction. Regardless of the estimation procedure

they used, it was found that mean WTP for the 5 in 1000 risk reduction was statistically

greater than mean WTP for 1 in 1000 risk reduction, which made their estimates pass the

scope test and proved sensitive to size of the risk reduction.

Another application of the CVM related to health care and services was carried out by

Belaineh Taye (2002) in Ethiopia. He estimated the WTP of 300 households in West Shoa

zone for insecticide-impregnated bed nets against malaria by using the open-ended

elicitation technique. He tried to present two scenarios to the households on how to acquire

the bed nets- in cash and in credit. His result shows that the WTP for insecticide

impregnated bed net in credit was higher (Birr65.05) than the WTP in cash (Birr 44.26).

Probably the first application of CVM for the evaluation of HIV/AIDS services anywhere

in the world is Forsythe’s PhD dissertation (2001) on the economic evaluation of

HIV/AIDS services in Kenya. Apart from evaluating the costs and benefits of four

HIV/AIDS services- voluntary counselling and testing (VCT), chronic care services, ARV

therapy, and a hypothetical AIDS vaccine- he assessed the factors that influence how

people value these services. In addition to these the researcher also tried to evaluate the

42
validity of CV in assessing benefits for HIV/AIDS programmes in developing countries. To

obtain the data, the study employed the payment card elicitation technique. WTP was

elicited using three survey instruments- VCT clients at the time of being counselled and

tested were asked about their WTP only for VCT; VCT clients six months after having

received their results were asked to state their WTP for chronic care services and ARVs

while in the third survey instrument, people within the general community were asked to

state their WTP for VCT, chronic care services, ARVs, an AIDS vaccine and their

willingness to contribute to an AIDS charity. Respondents were also encouraged to rank the

four services from highest community priority to the lowest. It was found that VCT was

given a high value intervention while AIDS vaccine was rated as the second highest

community priority. ART was ranked the least of all the health services. The findings of

Forsythe’s study indicate that PLWHA cited that they were willing to pay 11% of their

monthly household expenditures for ARV therapy. The amount of WTP obtained from the

data ($10.92 per month) was relatively smaller than the actual prices of the therapy ($210-

282 per month). The estimation thus indicates that in Kenya the price of ARV therapy

would need to be about $120 a year in order to be affordable for the government. With

regard to determinates of WTP it was found that household spending was closely related to

WTP responses across all four services. A focus on ARVs reveals that spending on

medications also had high correlation to WTP. The variables- number of visits to health

centre and having ever received chronic care service- were found to have weak correlation

to WTP for ART.

43
While Forsythe claims that his research is the first application of CVM for the evaluation of

HIV- related programmes, it is not the only one. Whittington et al. (2002) analysed the

private demand for HIV/AIDS vaccine in Guadalajara, Mexico. By considering a

hypothetical vaccine that would provide lifetime protection against HIV/AIDS, they

estimated the WTP of uninfected adults. The CV survey was administered to 234 adults

aged 18-60 years. The results of the estimates of private demand indicate that individuals

anticipate sizable personal benefits from such a vaccine, and that they would be willing to

allocate a substantial proportion of their income to be protected in this way from HIV

infection. It was also found in the research that individuals with higher incomes, with

spouses deceased or partners, and with perceived risks of becoming infected are WTP more

for the vaccine while older respondents are WTP less.

44
 CHAPTER FOUR

METHODOLOGY

4.1 Data source and type

The application of WTP in the areas concerning changes in individuals' economic welfare

and reductions in the probability of death (due to illness or accident) is based on

individuals' preferences. This presupposes that individuals treat longevity more or less like

any other good rather than as a hierarchical value. This economic approach for the

valuation of individual's WTP to increase his/her own life expectancy can be derived by

focusing on ex ante perspective- before the uncertainty about the individual's death during a

specified period of time is resolved; or ex post perspective- each individual knowing if s/he

is to die "now" or live a while longer. The former perspective is criticized for being morally

unacceptable (Freeman, 1993).

In this paper it is assumed that the availability of ARV drugs is most probably highly

valued by PLWHA than people without it. This is also attested by the findings of Forsythe

(2001) where HIV-positive clients had a higher WTP to have access to ARVs relative to

HIV-negative clients. Therefore, this study followed an ex post approach where by the

contingent valuation survey was used to solicit the willingness to pay for the drugs by

people living with the virus. Furthermore, instead of asking individuals from the society

whether they are HIV positive (especially due to stigma) and would pay for the drugs, only

45
residents of Addis who were willing to reveal themselves as carriers of the virus were

interviewed. This was administered through in-person interviews.

4.2 Sampling procedure

There are two national associations of PLWHA registered as NGOs to enable them to

operate. These are Dawn of Hope and Mekdim Ethiopia whose primary objective is to

protect the rights of members and to contribute to HIV/AIDS prevention activities. The

target population of this research is people living with the virus. The sample respondents

were drawn from the members Dawn of Hope and Mekdim Ehiopia. Attempt has also been

made to contact some medical doctors who have clients living with the virus.

From the above target group, sample was drawn by using opportunity sampling, which is a

method of non-probabilistic sampling. This is the simple expedient of including as subjects

whoever happens to be available from the population of interest. When members of the

respective associations contacted the association for some reason or another interviews were

conducted. Since the interviewers themselves were workers in the associations, it was also

possible to solicit responses when they visited members in their homes. At the same time,

clients with the virus of some doctors were also interviewed when they visited their doctors for

diagnosis. The total sample size of the survey was 440, which was obtained according to the

availability of people. A sample of 270 respondents from Dawn of Hope, 165 from Mekdim

Ethiopia and 5 patients from some doctors were included in survey.

46
4.3 Questionnaire Design

The CV survey begins with the description of the objective of the paper. This is followed

by socio-demographic questions. Part two of the questionnaire presented questions that help

to identify the knowledge, attitude and practice of respondents with respect to the drugs.

The next section of the questionnaire was the payment question, in which respondents were

presented with an initial bid amount and increasing (decreasing) amounts until the

respondent declined (or accepted) to pay. Respondents were asked if they would purchase

the drugs at $350 per year (approximately Birr 250 per month) or not. This threshold figure

is based on the current price of the drugs in the other parts of the world (UNAIDS, 2002

and Walgate, 2002).

However, since starting point bias occurs when the respondent’s WTP amount is specified

by a value introduced by the scenario, the paper used five different initial bid amounts (350,

300, 250, 200, 150) in order to take care of the “starting point bias” problem - an effect

related to the initial bid value. Those who responded “No” would be asked if they would be

willing to purchase the drugs at a lower cost. If they were still unwilling to buy the drugs at

all then they were asked to state or name the reason why they would not. On the other hand,

those who responded “Yes” were given a follow-up question in which s/he is willing to pay

a higher price (See Annex III for the copy of questionnaire used).

47
Figure 4.1. Bidding Game with Birr 350 to Birr 150 starting points.
Y Stop
Y bov e 5 Above5000
Y 501-600? NN Stop N
Stop Y ……..?
Y 400-500? N Stop
350? N Stop
N
Y Stop
. 250-3Y00? N Stop Y Stop Y Stop
. NN Y 51-100? 1-50?
, 151-200? N N Stop
. Y Y
. Y Above
.
Stop
Y 301-400? StY N
. Y 200-300? N Stop Stop
1.5150? N Stop 5000?
. N Y StopY Stop Y …..?
0 Y
… 51-100? Stop Stop
. N
T. he interview was conducted under two s1c-5e0n? arios: Satsosupming the drug lengthens life

years by indefinite years and ten years. This will also provide a test for consistency of

results with economic theory, in which WTP should be sensitive to the magnitude of

risk/mortality reduction.

Respondents then were asked some questions to identify the determinants of WTP. In

the last part of the survey questionnaire, respondents were asked about their attitude

towards government’s responsibilities.

All respondents were interviewed in Amharic. When translating the questionnaire

from English to Amharic great care was taken so as not to offend or irritate

respondents.

48
4.4 Methods of Data Collection

A total of twelve persons (10 enumerators and 2 supervisors) participated in the survey. All

of the twelve participants were workers in the two associations that live with the virus, a

deliberate action designed to avoid interviewer bias. This is so because the researcher

believed that it may be comfortable to respondents to be interviewed by a person who

knows how it feels to live with the virus.

A pilot survey was conducted by the researcher and some adjustments, such as including more

questions, re-ordering the questions and excluding offensive words were executed. Prior to the

main survey both enumerators and supervisors were trained. The fieldwork for the year 2003

started in the second week of February and was completed in the second week of March. The

excellent cooperation from workers of the associations, especially DOH facilitated the

fieldwork.

4.5 Method of Data Analysis: An econometric Approach

In the simple regression model:

Yi = β1+ β2xi+ui i = 1,2,…,n

We may observe the dependent variable only when Yi>0. However, the observed values of

the dependent variable may sometimes have a discrete jump to zero. A dependent variable

with the property that has a discrete jump at zero is known as a limited dependent variable.

When the dependent variable is limited by some value either from above or below, the data

is said to be censored (Chay and Powell, 2001). Suppose y* has a normal distribution, with
49
mean µ and variance σ2. We may consider a sample of size η (y1* , y2* ,…, yn* ) and record

only those values of y* greater than a constant c and for those values of y*≤c, we record the

value c.

The data is also said to be censored if the characteristics of respondents are available so that

the factors impacting on the decision not to be willing to pay for (for the drugs, in this

particular case), can be assessed. Some examples of censored data in empirical literatures

forwarded by Greene (2000) are: the number of extramarital affairs; the number of hours

worked by a woman in the labour force; the number of arrests after release from prison, etc.

Each of these studies analyses a dependent variable that is zero for a significant fraction of

the observations.

However, the first applications of the limited dependent variable model was given by Tobin

(1958) where he applied it to model household expenditure on automobiles (as cited in

Amemiya, 1985). Such models are referred to as Tobit models or as Censored

regressions.

Let's assume a utility maximization problem by defining the following symbols:

y = a household expenditure on a durable good

y0= the price of the cheapest available durable good

z = all other expenditures

x = income

50
Then the household maximum utility U(y,z) is subject to the budget constraint y+z ≤ x with

the boundary constraint y ≥ y0 or y = 0. Suppose y* is the solution of the maximization

subject to y+z ≤ x but not the other constraint, and assume that:

y* = β1+ β2x+u……………………………………….…(1)

Where u may be interpreted as the collection of all the unobservable variables that affect

the utility function. The solution to the original problem, denoted by y is:

y= y* if y* > y0
0 if y*≤ y0……………………….………..…(2)

Tobin (1958) assumed that y* is normally distributed and y0 to be the same for all

households. The standard Tobit model (Censored regression model) is then:

yi* = X'i β + ui i = 1,2,…,n,……………………..…....(3)

yi = yi* if yi* > 0

0 if yi* ≤ 0
Conventional regression methods fail to account for the qualitative difference between limit

(zero) observations and nonlimit (continuous) observations. Particularly, the standard OLS

regression using censored data will typically result in coefficient estimates that are biased

toward zero, neglecting the upward bias caused by the truncation.

The likelihood function for the Tobit is given by:

  Χi β  1  1 ( yi − Χ i β )2 
L= Π 1 − Φ σ  ⋅ Π exp −
σ2
 …………….(4)
   2Π σ 2  2 
yi / yi =0 i yi / yi >0 i

Where Ф = the distribution function of the standard normal

51
The first part of the likelihood resembles the probit while the second part resembles the

likelihood of conventional OLS on those sample points that are not censored (i.e., greater

than zero). The log-likelihood is:

 Χ β   1 1 ( yi − Χ i β ) 
2

l = ∑ ln 1 − Φ i  + ∑ ln −  ………………..(5)
yi / yi =0   σ  y i / y i > 0  2Π σ 2 2 σ2 

However, Greene (2000) and Johnston &DiNardo (1997) emphasize that if the underlying

disturbances are not normally distributed, then the usual estimator based on (5) is

inconsistent.

The coefficients of the Tobit model may not be sensibly interpreted. The expected value of

y in Tobit’s model is given by (McDonald and Moffitt, 1980):

Ey = Χβ F(z) + σ ƒ(z) ………………………………(i)

Where z = Χβ/ σ,

ƒ(z) = normal density

F(z) = the cumulative normal distribution function

Furthermore, the expected value of y for observations above the limit, y*, is simply Χβ plus

the expected value of the truncated normal error term

Ey* = E(y/y >0)…………………………………….(ii)

= E(y/u > - Χβ)

= Χβ + σ ƒ(z) / F(z)

52
Consequently, the basic relationship between the expected value of all observations, Ey,

and the expected value conditional upon the limit, F(z), is

Ey = F(z)Ey*………………………………………….(iii)

The McDonald and Moffitt marginal effect decomposition then is obtained by considering

the effect of a change in the ith variable of χ on y:

∂Ey  ∂Ey *   ∂F ( z ) 
= F ( z )  + Ey *   ………………………………………..(iv)
∂χ i  ∂χ i   ∂χ i 

The interpretation is that the change in the mean of y with respect to χi has two components.

One effect works by changing the conditional mean of y (the change in y of those above the

limit, weighted by the probability of being above the limit) and the other by changing the

probability that an observation will be positive (the change in the probability of being

above the limit, weighted by the expected value of y if above).

Powell's Least Absolute Deviations Estimator

Jim Powell (1984) proposed censored least absolute deviations (CLAD) estimator as an

alternative to maximum likelihood estimation of the parameters of the censored regression

model. This method does not impose normality and homoscedasticity on the distribution of

the disturbance term. In other words, it requires weaker assumptions on the error term than

the systematically trimmed estimator (Johansson and DiNardo, 1997).

53
Considering the standard index model:

y* = Xβ + ε

Where y* is not observed but rather y, where

yi = yi* if RHS >0

0 otherwise

which can be written as: yi = yi* if εi > - Xiβ

0 if εi ≤ - Xiβ

Powell notes that if y* were observed, and if the error term was systematically distributed

around 0, then standard OLS would produce consistent estimates of the parameters. Since

asymmetry into the distribution is introduced by censoring, for a given observation Xi , y*

is not observed and all observations where εi < - Xiβ are omitted.

If we assume that y* is observed in the basic censored regression model, then

E [yi*| Xi ] = Xiβ + E [εi | Xi ]= Xiβ…………………..………(6)

A consistent estimate can be obtained by OLS that is the solution to:

 n * ∧

min  ∑ ( yi − χ i β ) 2  …………………………………………….(7)
∧  i =1 
β

∧
That is, β is the estimator that minimizes the sum of squared errors. Instead, however, one

may choose to minimize the sum of the absolute value of the errors:

n * ∧ 

min ∑ yi − χ i β  ……………………………..…..……….(8)
∧  i =1 
β

which is called the least absolute deviations (LAD) estimator. This can be rewritten as

54
 n ∧ ∧

min ∑ ( yi − χ i β ) •ψ ( yi − χ i β ) …………………………….(9)
* *

∧ i =1
β

where the sign function ψ (•) takes on values of 1, 0, -1 (positive, zero, or negative). The

corresponding normal equation is given by:

n ∧
0 = ∑ χ i' • ψ ( yi* − χ i β ) …………………………………(10)
i =1

implying that it is the sign and not the magnitude of the residuals that matters. Therefore,

while OLS corresponds to mean regression, LAD estimator corresponds to median

regression that is consistent for β because

q50 [yi*| Xi ] = Xiβ + q50 [εi | Xi ]= Xiβ………………….(11)

In (11), q50 denotes the median of the fiftieth quantile.

Amemiya (1985) points out that Powell's estimator is attractive because it is the only

known estimator that is consistent under general non-normal distributions and that in large

samples, this estimator appears to perform much better than the various estimators of the

Tobit model.

This paper used both the Tobit and Powell's Censored Limited Absolute Deviation

(CLAD) to estimate the parameters of the data. In the latter case, the Bootstrap method is

used to estimate the standard errors. The Bootstrap due to Efron (1979) is a recently

constructed general technique for estimating sampling distributions (as cited in Jonston and

DiNardo, 1997). The bootstrap is believed to hold a great promise in many applications and

is finding its way into more and more applied econometric research. Researchers usually try

to determine the exact sampling distribution of the estimation procedure to compute

confidence intervals and for tests of significance. This, however, is often the hardest part of
55
the work and the reason for why asymptotic methods are quite often used for interval

estimation and hypothesis testing. The bootstrap provides a simple means for obtaining an

approximate sampling distribution of the statistic. This, however, is not as exact as the

exact distribution and is computationally longer to obtain results than by using the

asymptotic distribution. The bootstrap is preferred due to the fact that even if it is

computationally tedious, it is not computationally difficult and more importantly, the

results are conditional on observed data, not based on large sample approximations.

The formula for the standard error of the median is

smedian = 4 f 2 (0)

Where ƒ(0) is a consistent estimate of the value of the probability density function at 0. It is

possible to generate estimate of the distribution by using the data and then calculate ƒ2(0).

A second approach to calculating the precision of the median would be to

(i) draw a large number of samples of size n from the distribution ƒ(x)

(ii) calculate the median in each of these samples, and

(iii) calculate the square root of the variance of these estimated medians across a large

number of replications.

In both approaches it is possible to calculate a consistent estimate of the standard error if

one has a precise knowledge of the distribution generating the samples in the first place.

Since one does not know this, Efron suggested the use of the sample data to generate an

estimate of the distribution. That is, use the empirical distribution to learn about the actual

distribution. For a sample X and for i = 1,….θ the procedure is

56
 1. Generate a random sample Xi with replacement from the original sample X4.
∧ i
2. Compute the median M for this new sample.
∧ i
3. Store the value of M
Johnston and DiNardo (1997) underline that the number of bootstrap replications, θ, should

be set as high as is practical. The bootstrap standard error of the median is

2
1 n ∧ ∧

σ boot = ∑
θ − 1 i =1 
M i − M i (•) 


θ ∧
∧ 1
where M (•) =
θ
∑M i
i =1

4.6 Model Specification

The contingent valuation method was used to elicit the willingness to pay of people living

with the virus for antiretroviral drugs, which lengthen the lives of the particular individuals.

The median willingness to pay (MWTP) is specified as:

MWTP = β0+ β1SDV+ β2TC+ β3YLV+ β4HS+ + β5I+ β6XY+ β7WP+ β8BP+ β9FR

Where: SDV = Vector of Socio-demographic variables including age, sex,

marital status, religion, educational status, occupational status, family size and

family head.

TC = Total cost incurred for treatment of opportunistic diseases for the last
two months
YLV = Year(s) of living with the virus
HS = Current health status

4
We are drawing a sample of size n from the original sample, thus putting the probability 1/n on each
57
 I = Total income of the individual (personal, family, transfer)
XY = Payment for another person living with the virus in the family of the
individual
WP = Responsibility of payment of total treatment cost for the last two months
BP = Initial bid price bias
FR= Free riding

Table 4.1. The expected signs and relationships of the variables:

Variable Expected Explanation
Sign
Sex
Dummy variable Men usually are economically independent so that they
1 if male (+) have the decision power to pay for the drugs
0 female

Age ? The older an individual the greater are the family

responsibilities so that the higher the WTP for the drugs.
But the converse may not be true i.e., the younger the
person the less s/he is WTP for the drug because younger
people might be eager to live longer and achieve more
if they can make it. Hence the hypothesized
relationship of WTP and age is ambiguous.
Religion ? To be tested where r0 is treated as a control variable
r0= Protestant/Jehovah
r1=Orthodox/Catholic
r2=Muslim
Marital Status ? To be tested where m0 is treated as a control variable
m0= Single
m1=married
m2=divorced

observation in the sample.

58
m3=widowed
Educational Status (+) The more educated the individual is the greater
e0= Illiterate the awareness about the drugs’ advantages (e0 is treated
e1=Primary as a control variable)
e2=secondary
e3=post secondary
Employment Status (+) A person’s employment status, should be positively
Dummy variable correlated with her/his WTP if s/he is employed
1 if employed
0 Unemployed
? The greater the family size the higher is the consumption
Family Size
expenditure, other things being equal. Hence the
(FS)
individual is expected to have a lower WTP with
increased family size. However, the greater is also the
responsibility so that the higher the WTP.
. Hence the sign is ambiguous.
Family head The higher will be the WTP when the individual is the
(FH) (+) family head because s/he has to take care of the rest of
Dummy variable the family by lengthening her/his life
1 if head
0 otherwise
(+) The higher the income the greater will be the ability to
Total income pay for health facilities so that the greater will be
(I) the WTP for the drugs

The longer the individual lives with the virus the higher
Years of living will be the WTP because the individual is most probably
with the virus suffering and will be willing to do anything to stop it.
(YLV) ? On the other hand, those people who didn’t live with the
virus that long may not want to suffer for the their
remaining lives so that they may have a higher WTP
for the drugs.

59
Current Health Status If the individual’s current health status can be categorized
Dummy Variable( HS) under “good” or “bad” then s/he may have higher WTP
1 if bad (+) for the drugs when the health status deteriorates.
0 good
? The higher the treatment cost of different OIs for the last
Cost of Treatment two months the higher will be the WTP for the drugs
(TC) to shift the cost. However, Forsythe (2001) found a
negative sign where people who have already spent
a lot of money on medications, particularly medications
that are ineffective, are unwilling or unable to pay more
for ARVS. Therefore, sign is to be tested.
Who paid for (-) If it is already the responsibility of others to pay for the
treatment cost for the treatment cost of the individual for the last two months
last two months(WP) then this is an indication that s/he will likely have less
1 others willingness and ability to pay for the private good under under
0 self consideration.
Payment for another (-) If there is another person living with the virus and if
person living with it is the responsibility of the individual to pay for the
the virus in the drugs to that individual too then the less will be the
family WTP of the individual for the drugs.
(XY)
Dummy Variable
1 if yes
0 otherwise
Starting price of the
Bidding game ? To be tested
(BP)
Free riding ?
(FR)
To be tested
Dummy Variable
1 if others pay
0 otherwise
60
 Chapter Five

EMPIRICAL FINDINGS AND ANALYSIS

In this chapter, results of the CV survey are presented and analyzed using both descriptive

and multivariate regression analysis. For the descriptive analysis SPSS 11:0 is used while

the STATA software (version 7.0) is used to run Tobit and CLAD to estimate the

parameters that determine WTP for ARV drugs.

5.1 Descriptive Analysis4

In this section, the CV survey results are summarized in six sub-sections. The first part

presents the socio-demographic characteristics of respondents followed by questions

designed to capture their knowledge, attitude and practice (KAP) of the drugs. The third

section of the analysis is on the WTP amounts of the respondents. The next two parts

4
Refer Annex 1 for summary tables
61
concentrate on the economic characteristics of respondents and health related indicators.

Finally, the general attitudes of respondents are summarized. The remaining three sub-

sections focus on different tests of the CV method employed in the survey.

5.1.1 Socio-Demographic Characteristics

Data on the gender of the 440 respondents shows that 264(60%) are female out of which

91(72.3%) are family head while 176(40%) are male respondents and 134(76.1%) of them

are family heads.

The mean age of the respondents is 33 years. The youngest respondent is 18 years old while

the oldest is 65 years old. The total number of respondents in the age range 18-49 years is

426 (97%). This reflects the fact that around 91% of infections in the country occur among

economically active age group (15-49 years). The proportion of respondents within the age

group 18-25 years old is 80 (18.2%). The majority of the respondents, 167(38%), fall into

the age range 26-33 years followed by 134(30.5%) of respondents between 34-41 years.

Those between 42-49 years are 45(10.2%). Only 12(2.7%) and 2(0.5%) of the respondents

are among the age group 50-57 and 58-65 years, respectively.

The average family size of the respondents is 3.85, ranging from 1 to 12. Out of the 440

interviewees, 205 (46.6%) have 1 to 3 family members, which suggests a somewhat smaller

family size among the majority of the respondents. Family size of 4-6 individuals is

reported by 191(43.4%) of respondents and 7-9 by 41(9.3%) of respondents. Of the total

respondents, only 3(0.7%) individuals reported a family size of 10-12.

62
Data on marital status reveals that 170(38.6%) respondents are currently married, which is

followed by those whose spouses are deceased, 132(32%). The rest of the respondents

84(191%) and 54(12.3%) are either divorced or single, respectively.

Most of the respondents, 282(64.1%), indicated that they are Orthodox Christians. Eighty-

two (18.6%) of the respondents are Protestants and 65(14.8%) are Muslims. The rest of the

respondents indicated that they are either Catholic (1.4%) or Jehovah (1.1%).

In terms of educational status, most of the respondents, 196(44.55%), have completed high

school followed by 102(23.2%) respondents that have completed primary education. Those

that have beyond secondary education level are 64(14.5%). Twenty-eight (6.4%)

respondents indicated that they have only below 4th grade training. The rest of the

respondents, 30(6.8%) and 20(4.5%) are either illiterate or able to read and write,

respectively.

The last variable that is considered in this sub-section is the occupational status of

respondents. Those who claimed to have their own business account for 17.3% of the total

respondents where as those that work as private employees and civil servants are 55

(12.5%) and 33(7.5%), respectively. NGO workers constitute 16.1% of the total

respondents. Thirteen respondents (3%) were workers of the respective associations

established by PLWHA while 31(7%) indicated that they are daily labourers. However, the

majority of the respondents, 161(36.6%) are found to be unemployed during the survey.

5.1.2 KAP
63
In this section the knowledge, attitude and practice of respondents with regard to ARV

drugs are summarized.

Interviewees were asked if they ever heard about ARV drugs before the day of the

interview. Even though almost all, 99.5%, confirmed that they have heard about the drugs

before, only 28(6.4%) of the respondents are or have been using the drugs purchasing from

the parallel market.

Out of the 440 respondents, 177(40.2%) reported that they know people that use the drugs

among friends, relatives or those living around.

Respondents were asked if they knew or heard about the current ‘underground’ price of the

drugs in the country and if they thought this was reasonable. Those who think that the

current price is not fair compared to the living status of the population are 99.5% while

2(0.5%) of the respondents admitted that they do not know about the on going price in the

country.

To evaluate the perception of respondents towards the drugs, they were asked to give their

opinion about for how long the drugs extend the life of a person living with the virus. More

than half, 243(55.2%), of the respondents either do not have a clue or claim that only God

knows. Among those respondents that guessed the years, 11(2.5) believe that the drugs

extend life years from 2 to 3 years; 79(18%) think that it extends from 5 to 10 years;

44(10%) conjecture from 10 to 15 years; 34(7.7%) guess from 15 to 20 years. Out of the

total respondents, 8(1.8%) and 3(0.7%) respondents speculated that the drugs lengthen life

years from 20 to 25 and 25-30 years, respectively. The rest 18(4.1%) respond that once the

drugs are started to be taken by an HIV+ person, life is elongated indefinitely.

64
Respondents' perception about the side effects of these drugs was also used as a means of

evaluating their knowledge regarding the drugs. The response to this particular question

ranges from 'no toxicity' to 'very high toxicity'. However, 233(53%) let us know that they

do not have a clue about the side effects of the drugs. Those who think the side effects to be

'modest' and 'low' are 66(15%) and 64(14.5%), respectively. The two extremes, 'very high

toxicity' and 'no toxicity' are chosen by 28(6.4%) and 49(11.1%), interviewees,

respectively.

5.1.3 Willingness to Pay

In spite of their ignorance or speculation with regard to for how long the drugs lengthen life

years of PLWHA once they start taking the drugs, respondents were made to assume two

different scenarios- the drugs lengthen life years indefinitely and ten years.

Out of the total 440 respondents, 242(55%) are either unable or unwilling (protest votes) to

pay for the drugs. The majority, 210(47.7%) respondents, indicated that they do not have

the ability to pay for the drugs. The other 32 respondents appear to have “protest votes”

where their responses generally are recognized not to reflect the true worth of the service,

but rather a protest against the idea of consumers having to pay any amount for a certain

good or service (Mitchell and Carson, 1993). Out of these 26(5.9%) claimed that there is no

chance they would pay for the drugs because only God can extend their life years. The rest

4(0.9%) and 2(0.5%) respondents are unwilling to pay because they believe that the drugs

65
should be given for free or because of the side effects, respectively. Hence WTP responses

in this descriptive analysis are analyzed from 198(45%) of all respondents.

a) Mean Vs Median WTP

The minimum stated price that respondents are willing to pay is Eth. Birr 50 while the

maximum is Birr 2000.

The mean WTP of respondents if the drugs lengthen life years indefinitely is Birr 163.64

while it is Birr 148.48 if the drugs are effective only for 10 years. On the other hand, the

median WTP of respondents if the triple combinations lengthen life years indefinitely is

Birr 100 whereas it is Birr 50 if the drugs are going to extend life years by only 10 years.

One of the statistical issues under the analysis of WTP is to consider whether to use the

mean or median WTP estimates. In this study the median WTP is used to estimate the WTP

of PLWHA for ARV drugs and the determinants of their WTP amounts. This is due to the

fact that the mean is very sensitive to assumptions about the valuation function, skewness

in the distribution and to outliers in the data, which our data is no exception (refer Annex 2)

whereas the median is more robust in these cases.

If the market price of ARV triple combination per month is to be Birr 1-50, 105(53%) and

92(46.5%) of the respondents are willing to pay when the expected life extension is 10 and

indefinite years, respectively. When market price ranges from Birr 51-100 per month, it is

chosen by 40(20.2%) and 44(22.2%) respondents when the expected life years extension is

10 and indefinite years, respectively. Nine(4.5%) and 16(8.1%) respondents are willing to

pay from Birr 101-150 per month for the drugs if life extension is 10 and indefinite years,

66
respectively. Willingness to pay for the drugs if market price is between Birr 151-350 is

observed in 21(10.6%) respondents if the drugs extend life years by 10 years and in

19(9.6%) respondents for a 10 years life extension. The reverse is true when the market

price for triple combinations is from Birr 351-550 i.e., 19(9.6%) respondents willing to pay

for 10 life years extension and 21(10.6%) respondents if it is indefinitely. In the last

category of WTP, Birr 551-2000, only 4(2%) and 6(3%) respondents out of 198 individuals

who showed interest to pay are willing to pay the above amount if the drugs lengthen life

years by 10 and indefinite years, respectively.

b) WTP and Starting Bid

To obtain the WTP of PLWHA for ARV drugs, five

different initial bid prices were given to different
respondents randomly. The prices were Eth. Birr 150,
200, 250, 300 and 350 per month and 91, 110, 79, 82,
and 78 individuals were randomly given the respective
starting prices. Out of the 91 respondents given Eth.
Birr 150 as a starting price, 15.7% respond yes and the
rest no. Out of the 110 individuals that were given 200,
29.29% are willing to pay for the drugs. Those that
were presented with Eth. Birr 250 per month
responded yes. For the fourth starting bid price,
26.77% of respondents out of the 82 individuals that
were given the same initial bid showed interest in the
payment for the drugs. Finally, those 78 respondents
provided with an initial price of Birr 350 per month,
16.16% showed willingness to pay for the drugs.
c) Payment responsibility

67
Out of the total 198 respondents that stated their willingness and ability to pay for the

drugs, 73.24% responded that they themselves (60.6%) or their families (12.6%) are going

to pay while 26.8% stated that it is the responsibility of others- either their respective

associations (21.7%) or the organizations they work in (5.1%).

d) WTP and Some Socio-Demographic Characteristics

In this section an attempt to cross-tabulate some

selected socio-demographic characteristics that are
thought to strongly influence the decision of WTP is
made .
The cross tabulation of WTP with gender of
respondents reveals that when the price of triple
combinations is very low, the percentage of women who
are WTP is greater than men. However, when the
figure increases, the percentage of male respondents
who are WTP the respective figures, is greater than
female respondents.
The minimum WTP amount, Birr 1-50 and the
maximum WTP amount, Birr 551-2000 are associated
with mean ages 32.29 and 38, respectively. The other
WTP amount categories are associated with mean age
ranging from 37.31 to 37.95 years.

The relationship between WTP and mean family size

reveals that the maximum amount of WTP (1500-2000)
is associated with mean family size of 4 while the
minimum WTP amount (1-50) is related to mean
family size 3.85.
68
Concerning the occupational status of respondents, for the WTP amount between Birr 1-50,

the majority of the respondents are unemployed. The fact that being unemployed does not

necessarily mean that the respondents have zero income level because the survey considers

total income of an individual: not only income earned per month by the individual but also

family and transfer income. Most of the respondents that showed interest in paying the

second interval of the WTP amount, work in different NGOs. This is also true for the WTP

category Birr 151-350. Those that have their own business are the majority that showed

interest to pay from Birr 351 to Birr 2000.

5.1.4 Economic status

To evaluate the total income of respondents per month, three sources were identified- the

respondent’s earning, family income and transfer income. The latter refers to money

income transferred from the respective associations, relatives or friends to the individual.

The mean income of the respondents is Eth Birr 367 per month.

Out of the 440 respondents, 144(32.7%) of them have a total income below Eth. Birr 120

per month. Respondents that get total income from Birr 150 to 250 per month are

78(17.7%). This range is almost mated by the next income figure, Birr 300–400, where

77(17.5%) of the respondents get per month. Out of the total interviewees, 50(31.4%) and

33(7.5%) of the respondents fall into income group Birr 450-550 and Birr 600-700 per

month, respectively. Only 11(2.5%) of the respondents have total income Eth Birr 750-850

per month. In the last category of income, 47(10.7%) of the respondents get above Birr 900

per month.

69
The mean income of the 198 respondents that stated different willingness to pay amounts is

Birr 583. Comparing this with the mean WTP indicates that these people are prepared to

allocate 28% of their income if the drugs are effective indefinitely and 25% if only 10

years.

4.1.5 Health related indicators

In the fourth section of the questionnaire, respondents

were asked for how long they knew they were living
with the virus. This ranges from 15 days to 13 years.
Out of the 440 respondents 99(22.5%) reported that it
is less than one year since they knew that they were
living with the virus. The majority of the interviewees,
147(33.4%), reported from 1-2 years; 95(21.6%) from 3
to 4 years; 43(9.8%) from 5 to 6 years; 29(6.6%) from 7
to 8 years; 20(4.5%) of the respondents claimed 9 to 10
years. For 5(1.1%) of the respondents, it has been 11-12
years since they knew they were living with the virus.
Out of the 440 respondents, 2(0.5%) individuals lived
the longest with the virus, that is, 13 years.
To evaluate the health status of respondents three
approaches were followed. The first was self-
evaluation, that is, by giving the chance to respondents
themselves to rank their health status. Then
respondents were asked for how many times, for the
last two months, did they visit a hospital, clinic or
health station. Since some people may prescribe anti-
pain for themselves without consulting a doctor, the
70
third approach was to ask respondents whether they
bought medicines from pharmacies without
prescription from their doctors. The responses for the
above three types of questions were used to categorize
the health status of respondents.
The data for health status show that most of the respondents, 139(31.6%) are in a ‘good’

health condition and 116(26.4%) in ‘modest’ health status. Some 114(25.9%) respondents

think that their health status is very good and they neither visited a doctor nor prescribed

medicine for themselves for the last tow months. These are categorized as having ‘very

good’ health status. Out of the total respondents, 59(13.4%) of them have ‘bad’ health

status and only 12(2.7%) show to have ‘very bad’ health condition.

The cross tabulation of the above two variables in this sub-section reveals that, out of the

99 respondents that knew they were living with the virus for less than a year, the majority,

31respondents, have ‘good’ health status. The same is true for those who have known that

they are HIV+ from 1 to 2 years and 3 to 4 years. However, those individuals who

responded that they knew for 5 to 6 years and 7 to 8 years, their health status is categorized

under ‘modest’. Out of the 20 individuals that have been living with the virus for 9 to 10

years, most of them, 35%, have ‘bad’ health status. Of the 5 respondents that knew about

the virus they are living with for 11-12 years, 4 of them have ‘very bad’ health condition

and both of the respondents that reported to be living with the virus for 13 years also have

‘very bad’ health status.

71
From the total respondents in the survey, 294 of them reported to have incurred treatment

cost for the last two months. These respondents were asked how much they spent and who

paid for the medical expenses. The average total cost is found to be Birr 255, which is 64%

of the respective respondents’ income.

Respondents that spent from Birr 1-50 are 103, out of which 26(25.2%) paid out of their

pocket; for 7(6.8%) respondents their families covered the cost, for another 7(6.8%)

respondents the organization they work in paid for the medical cost. However, for the

majority, 63.2% of the respondents, the respective associations covered the treatment costs.

Those respondents that claimed to incur a medical cost of Eth Birr 100-200 for the last two

months are 83 and again for the majority, 52(62.6%) of the respondents, the cost was

covered by the associations whereas for 24(30%) respondents either the individuals

themselves or their families paid for the treatment cost. For the rest 7(7.4%) individuals that

incurred the above figure, the organizations that they work in paid for treatment. Of the 69

respondents that reported to pay from their pocket, 11(16%), of them paid Birr 250-350 for

the last two months for medical treatment; 7(10.1%) around Birr 440-550; only two

respondents paid Birr 550-650 and another two above Birr 700. Of the total respondents

that incurred medical cost for the last two months, 181 respondents’ costs were covered by

the respective associations. Out of these, the associations paid Birr 250-350 for 30(16.6%)

respondents; Birr 400-500 for 19 individuals; and Birr 555-700 for four respondents.

Data related with the issue of another HIV+ person living with the respondents reveal that

156(35.5%) of them actually have a person/s living with the virus at home. Of these

79(50.6%) are the spouses of the respondents; 47(30.1%) constitute the child/children of

72
the respondent; 10(6.4%) relatives living with the respondents in the same house;

18(11.5%) of the respondents who claimed to be living with other HIV+ persons, the

persons are his/her spouse and child/children. The rest 2(1.3%) out of the 156 respondents

live with both their HIV+ child/children and relative/s.

However, even if 156 of the respondents responded that they live with another HIV+

person, only 26(16.7%) showed willingness to pay for the other person/s if the drugs

become available in the market whereas 130(83.3%) of them refused to have this

responsibility.

5.1.6 General Attitude

All of the 440 respondents were asked to give their opinion about the responsibility of

distributing the drugs. The majority of respondents, 215(48.9%), believe that government

has the responsibility of distributing the drugs for free; 52(11.8%) said it should be left to

the NGOs operating in the country to dispense the drugs for free; 155(35%) asserted that

both government and NGOs should jointly distribute the drugs for free. Around 5(1.1%)

respondents urged their respective associations to take the responsibility of giving out the

drugs to their members. Those who suggested that, on the one hand, government should

distribute the drugs for free for those unable to pay and, on the other hand, private

pharmacies should sell to those that can pay are 11(2.5%). Two individuals out of the 440

respondents said that government, NGOs and the associations must work together for the

free distribution of the drugs.

73
Respondents were inquired if the Ethiopian government has given proper attention till the

day of the interview. Of the total 440 interviewed, 396(90%) disagree (most of them

strongly) with the suggested idea while the rest 44(10%) respondents believe that it is doing

its best concerning ARV drugs.

The last question presented to respondents was if they would be interested to pay more than

what they have stated before if their income increases and their economic situation changes

for the better. The majority, 408(93.5%), of the respondents confirmed this while 32(7.3%)

stated that no matter what their economic situation is, they are not WTP for the drugs. The

reason given by the majority, 26(81.3%) of them is only the Lord could extend their life

years and not the drugs. The rest four and two respondents are totally unwilling to pay for

the drugs because they believe either it should be given for free or it has high toxicity. This

question has helped the researcher to check whether there is internal consistency in the

responses of interviewees. For instance, it would have been contradictory on the part of the

respondent if s/he gave a reason for the protest vote as “only God can help me” in section

three but then assert that s/he would pay a positive amount if her/his economic situation

changes for the better.

5.1.7 Validity Test

To test the accuracy of the CV survey of this study, construct and scope validity tests are

carried out. Construct test refers to the theoretical validity test, which involves assessing the

degree to which the findings of a study are consistent with theoretical expectations

(Mitchell and Carson, 1993). The most common test for construct validity of CV studies is

the correlation between willingness and ability to pay (Forsythe, 2001). This is so because
74
if people were overbidding without giving any thought to their economic limitations, one

would expect that there would be no correlation between income and WTP.

The data of the survey reveal that there is a significant level of correlation between income

of the respondents and their WTP (Table 5.11). The correlation coefficient between total

income of respondents and their WTP for the drugs per month life is extended indefinitely

is 0.63 whereas the correlation coefficient of total income and WTP if the drugs extend life

years by 10 years is 0.64. Both of the Pearson correlation coefficients are significant at 0.01

level of significance.

Another test of the validity of a CV study is to test if respondents can distinguish the

intervention according to the scope of the intervention programme (Forsythe, 2001). In

order for WTP to provide accurate estimates, respondents should be WTP more for an

intervention that has a larger impact. In other words, they should place a higher value on

the drugs if these lengthen life years indefinitely than only ten years.

Respondents have apparently passed the scope sensitivity test because their WTP for ARV

drugs when life years are extended indefinitely is greater than when the effectiveness is

only ten years, which is consistent with economic theory stated in the third chapter i.e.,

WTP should be sensitive to the magnitude of mortality/risk reduction and should be higher

for higher mortality/risk reductions.

5.1.8 Starting Point Bias Test

Studies whose WTP amounts are highly correlated with the starting points are so biased as

to be invalid (Mitchell and Carson, 1993). Pearson correlation coefficients of the two WTP
75
amounts were tested (Table 5.8) and the result shows that the correlation coefficient

between the WTP for indefinite years extension and starting point is 0.024. Similarly

Pearson correlation coefficient between WTP for ten years extension and starting point is

0.044, which both of the coefficients are insignificant. Therefore, this descriptive analysis

suggests that the CV survey may not suffer from starting point bias.

5.1.9 Free Riding Test

Respondents may over pledge their WTP amount if they believe that others are going to

pay for the intervention under consideration. To test whether this CV survey suffers from

free riding problem, comparison between the population median WTP and the median WTP

of those who claimed that it is either their organizations’ or associations’ responsibility to

pay for the drugs is made. If the latter is far higher than the former it is to be concluded that

there is in fact free riding on the part of respondents.

As can be seen from Table 5.10, the median WTP is Birr 50 for both indefinite and ten

years scenarios, which is not actually far from the median WTP of the population. Hence, it

is possible to conclude in this descriptive analysis that the CV survey carried out for this

research does not suffer from free riding problem.

5.2 Regression Analysis

This sub-section presents results of both parametric and semiparametric regression

techniques and property of the data used in the regression analysis. The determinants of
76
WTP for ARV drugs by PLWHA were estimated with a sample size of 408 records (after

dropping 32 protest values).

5.2.1 Parametric estimation of the censored regression

Table 5.12 presents results of the parametric estimation of the censored regression using the

collected data. The model estimates determinant parameters of WTP for ARV drugs

assuming that 210 observations are left-censored (at zero) and 198 uncensored

observations. The pseudo R2 is 0.28 indicating that 28% of the variation in WTP is

explained by the independent variables identified in the study. The likelihood ratio for the

estimation model indicates that the overall model is a good fit.

As stated in the previous chapter, to meaningfully analyze the magnitude of the effects of

the regressors on the dependent variable, it is necessary to analyze their marginal effects.

This involves decomposing the unconditional mean into the effect on the probability of

WTP and the effect on the conditional level of WTP. For continuous variables in the model

these marginal effects are used to calculate elasticities at the sample means to allow a

quantification of the magnitude of the effects to be made. The effect of the discrete or

categorical variables on the dependent variable, however, cannot technically be quantified

using elasticities. Instead, marginal effects are used to calculate percentage changes in the

dependent variable as a result of the variable moving from zero to one, ceteris paribus

(Newman, 2001).

77
Table 5.12 Results of the Tobit Estimation

Coefficient dF/dx dF/dx

Variable (conditional) (probability) Z
0.6865141 0.273953
Employed^ 1.985436 (0.14) (0.06) 4.83***
-0.1198156 -0.0542644
FS -0.4085387 (0.03) (0.01) -3.83***
0.5758947 0.2608222
I 1.963645 (0.04) (0.02) 15.53***
0.0926122 0.041944
TC 0.3157825 (0.04) (0.02) 2.1**
-0.3768145 -0.1691682
WP^ -1.286819 (0.13) (0.06) -2.94***
0.1037006 0.0469659
BP 0.353591 (0.04) (0.02) 2.5**
-2.22245 -1.006546
Cons -7.577952 (0.38) (0.17) -5.78***
No. of observation = 408 Obs. summary: 210 left-censored obs.
Log likelihood = -517.25203 198 uncensored obs.
LR χ2 (21) = 402.99
Prob > χ2 = 0.0000
Pseudo R2 = 0.2803

^ dF/dx is for discrete change of dummy variable from 0 to 1

* Significant at 10%, ** significant at 5%, *** significant at 1%

The result in the above table reveals that apart from employment status and family size, all

other socio-demographic variables are found to be insignificant in this parametric

estimation. When a person living with the virus is categorized as employed, the WTP

increases and this variable is found to be statistically significant at 1% level. The marginal

coefficients indicate that an employed person increases the expected value of WTP by

68.7%, holding all other variables constant. The probability that an employed person will

be willing to pay more is 0.27 than an unemployed one.

78
The estimated elasticity coefficient for family size is not only negative, as hypothesized,

but also is statistically significant at 1% level. The marginal coefficient indicates that an

additional member in the family would decrease the WTP of the individual by 12%.

The finding of this parametric estimator shows that income elasticity is highly significant,

at 1% level of significance. As hypothesized, the coefficient’s sign is positive. This is in

line with economic theory that states that income has a direct effect on demand for normal

goods and services. Hence, a 1% increase in income will increase WTP of the individual by

58%.

The elasticity coefficient of total treatment cost for the last two months reveals that the

higher the cost, the higher would also be the WTP of the individual, which this variable is

found to be statistically significant at 5% level. The reason might be that PLWHA would

like to shift the cost of treatment to ARV drugs rather than treat opportunistic infections.

Thus, a 1% increase in the treatment cost of OIs will result in a 32% increase in the WTP of

the individual for antiretroviral drugs, holding all other variables constant.

The fact that other parties (organization or association) pay for the treatment cost for the

last two months, instead of him/herself, implies a decrease in the expected value of the

WTP by 37.7%, ceteris paribus. Again others cover for treatment cost also means that the

probability of the individual to pay more declines by 0.17.

The explanatory variable, initial bid price was used as a test for starting price bias. The

result shows that this variable is significant at 5% level of significance. However, free

79
riding is identified to be not a problem of the study. This is in conformity with the result of

the descriptive analysis, unlike starting bid price bias.

5.2.2 Data Property2

Before dwelling into the discussion of the result of censored limited absolute deviation,

exploration of the CV survey is made to identify the properties of the data.

Mitchell and Carson (1993) underscore that contingent valuation surveys are particularly

susceptible to outliers because WTP amounts, unlike ordinary scales, are unbounded at the

upper end and this survey is no exception. Even if it is possible to delete outliers on an ad

hoc basis, Mitchell and Carson (1993) argue that this procedure suffers from the obvious

drawback that it opens the researcher to criticism. Instead the authors suggest to use robust

statistical estimators in order to mitigate the effect of outliers.

An attempt was also made to test whether the data of the survey encounters econometric

problems i.e., non-normality, heteroskedasticity and multicollinearity.

The two useful parameters, skewness and kurtosis, are used to characterize the normality of

the distribution. Skewness coefficient is used to measure the symmetry of a distribution

while kurtosis measures the thickness of the tails of the distribution (Greene, 1997).

The test for these coefficients that is carried out for continuous variables in the data shows

that WTP, total income, treatment cost and years of living with the virus prove non-

2
Refer Annex 2 for attached results
80
normality since the skewness and kurtosis coefficients of these variables are different from

zero and three, respectively.

To detect multicollinearity, a simple correlation coefficient matrix of independent variables

was used. However, the result shows that multicollinearity is not a serious problem of the

data because a correlation coefficient between any of two regressors is not found to be as

high as 0.8, which is the rule of thumb for the detection of multicollinearity (Gujarati,

1995).

A Cook-Weisberg test3 for heteroskedasticity was carried out. Using fitted values, the test

result shows that the null hypothesis of constant variance is rejected since the calculated χ2

at one degree of freedom obtained from the estimated model is 557.65 with prob> χ2 =

0.0000. This implies that there is heteroskedasticity problem in the model. This, however, is

a common problem, which arises in the analysis of cross-section data.

Therefore, non-normality and heteroskedasticity are detected in the data. This fact coupled

with the result of the descriptive ananlysis, where 242 observations are censored at zero,

necessitate the use of censored quantile regression (CLAD). This is so because, unlike the

standard estimators of the censored regression model such as Tobit or other maximum

likelihood approaches, the CLAD estimator is robust to heteroskedasticity and is consistent

and asymptotically normal for a wide class of error distributions.

5.2.3 Semiparametric estimation of the censored regression

81
In this section, results of the censored least absolute deviation (CLAD) method of Powell

(1984) are presented. The estimation model proceeds by minimizing the (positive) sum of

deviations, rather than maximizing the (negative) sum of loglikelihoods.

The estimation procedure followed in this paper is the one outlined by Johnston and

DiNardo (1997) i.e.,

1. Run LAD4 on the entire sample to generate an initial estimate of β.

2. Use this estimate of β to drop observations for which the predicted value is

negative.

3. Run LAD on this new sample, to calculate a new estimate of β.

∧
4. Repeat steps 2 and 3 using the β in step 3 as the new initial estimate.

5. Continue until the estimate stops changing.

As discussed in the methodology part, CLAD estimates the standard errors using the

bootstrapping method. Johnston and DiNardo (1997) propose that the bootstrap replications

should be set as high as is practical. In this study the replication is set to 1000.

Even if it is not uncommon to find low R2 results from CV studies, Mitchell and Carson

(1993) note that CV studies are open to question if the R2 fails to show at least 0.15. The

pseudo R2 from the CLAD estimation is 0.339, which makes this particular study to be

reliable. This result is an improvement over the parametric estimation, which is 0.28.

3
This test was carried out using hettest stata command by first estimating the data using OLS
4
The regression command bsqreg is used
82
Unlike the parametric estimation result, age of respondents is found to be significant at

10% level. The sign of this explanatory variable is positive indicating that as age increases,

the tendency to pay also increases.

With respect to marital status, compared with unmarried people both estimation techniques

show that those that are married, divorced or whose spouses deceased have lesser WTP.

CLAD estimator particularly indicates that the latter is significant at 5% level of

significance. The explanation might be that most probably these people lost their spouses

due to the virus and have to support the remaining members of the family, which obviously

increases the burden of responsibility. In addition, due to the high treatment cost for the late

spouse, the widower/widowed may have exhausted his/her resources. Hence, their WTP is

lesser than singles.

Family size, in both parametric and semiparametric estimation techniques, is found to be

negative and statistically significant at 1% level. One explanation is the greater the family

size the higher is the consumption expenditure in which case the individual is expected to

have lower WTP for the drugs.

Income also exhibits consistency in both estimators- positive and highly significant

compared with the other explanatory variables at 1% level of significance.

83
Among the health status indicators, total cost for the last two months is found to be positive

and significant, but in this estimation it is at 10% level, instead of 5% as in the case of the

Tobit estimation.

WP, the responsibility of treatment cost coverage, also exhibited different level of

significance (5% level of significance instead of 1%) with the sign of the coefficient being

unchanged.

The initial bid price is found to be positively and weakly affecting (at 10% level of

significance) the final response to WTP questions.

Both the parametric and semiparametric estimation methods found free riding as not a

problem of the study.

As can be seen in Table 5.13 among the socio-demographic variables that are statistically

insignificant, the sign of sex is negative in both the Tobit and CLAD estimators suggesting

that female respondents are more WTP than male respondents. A priori the contrary was

hypothesized. Another variable, religion, was also found to be insignificant. This is in line

with Forsythe’s result (2002) that found religion to be an insignificant determinate of the

WTP for ARV therapies. With regard to educational status, compared to illiterate

individuals,those that have primary and secondary level of education are more WTP for the

drugs. Surprisingly, both the parametric and semiparametric methods show that compared

with illiterate people, those who have post-secondary education have less WTP. Even

though this variable is insignificant, it needs further investigation. Employment status was

84
found to be significant by the parametric estimation technique but not in the semiparametric

method though the sign is preserved.

Table 5.13 Result of CLAD estimation

Variable Coefficient Std. Err. T P>t

Sex -0.4241351 0.5119112 -0.83 0.409

Age 0.0727028 0.0367028 1.98** 0.05
Marital status
Married -0.4736836 0.5319854 -0.89 0.375
Divorced -0.9356687 0.5913206 -1.58 0.116
Spouse deceased -2.312446 0.9379913 -2.47** 0.015
Religion
Orthodox/Catholic -0.7729743 0.6022715 -1.28 0.202
Muslim -0.0853353 0.8674373 -0.1 0.922
Education
Primary 0.8700452 0.9702863 0.9 0.372
Secondary 0.3779187 0.8472433 0.45 0.656
Post secondary -0.442761 0.9326168 -0.47 0.636
Employed 0.9709493 0.8453755 1.15 0.253
FS -0.5571477 0.1306616 -4.26*** 0
FH 0.4900646 0.5995701 0.82 0.415
I 1.939576 0.2727556 7.11*** 0
YLV -0.079965 0.1005682 -0.8 0.428
HS -0.1224018 0.9272038 -0.13 0.895
TC 0.732034 0.4247543 1.72 0.086*
WP -1.033163 0.5036821 -2.05** 0.042
XY 0.1503784 0.9928933 0.15 0.88
BP 0.3224441 0.177465 1.82* 0.072
FR -0.4194028 0.4783331 -0.88 0.382
_cons -7.594917 2.012666 -3.77*** 0
Initial sample size = 408
Median regression, bootstrap(1000) SEs Final sample size = 151
Raw sum of deviations 386 (about 2)
Min sum of deviations 255.1433 Pseudo R2 = 0.3390
85
Among health status indicators that are found to be
statistically insignificant, years of living with the virus
has a negative sign indicating that as the years of living
with the virus increase the WTP for the drugs declines.
The same is true for the current health status of the
respondents.

Some of the explanatory variables changed their signs

when CLAD was run. These are health status, family
head, responsibility of payment for the drugs for
another person living in the family, and free riding,
which all are insignificant in both estimation
techniques.

To examine whether the effects of the significant

explanatory variables in the above median distribution
still hold, the 25th and 75th quantile regression were
run, in addition to the 50th quantile regression (Annex
1, Table 5.14). The psuedo R2s for the 25th and 75th
quantile regression are 0.27 and 0.49, respectively.

For the 25th quantile regression, family size, total cost

of treatment and initial bid price were found to be
statistically significant at 5% level where as income is
still significant at 1% level. The responsibility of
payment for the treatment cost was found to be
significant at 10% level of significance. All of these
statistically significant variables have the same signs as
the 50th quantile regression.

86
On the other hand, the 75th quantile regression showed
that employment status, like the Tobit result, is
significant but this time only at 10% level. Family size
and income are also significant in this estimation at 1%
level of significance and family head at 10%.
Post secondary education has positive sign across these
new quantile estimations implying that compared to
illiterate people, respondents having this level of
education are more WTP for ARVs. However, all of the
educational status variables are still insignificant. In
both the 50th and 75th quantile regressions, being a
family head implies more WTP whereas the 25th
quantile estimation shows the opposite effect. Another
variable that changed its sign, from negative to
positive, is years of living with the virus. In both the
25th and 50th quantile regressions, the longer the years
of living with the virus, the lesser the WTP of
individuals whereas the effect of this variable is direct
on the dependent variable when the 75th quantile
regression is run.

5.2.4 Summary of Estimation Results

The estimations of the data of this paper empirically assessed the determinants of the WTP

for ARV drugs by PLWHA.

Apart from standard parametric model used on censored data, notably the Tobit model, a

semiparametric estimation technique is also employed, which makes less restrictive

assumptions about the functional form of the disturbance term. The results of both

techniques are contrasted.

87
In the Tobit estimation, employment status and family size were found to be determinants

of WTP among the stated socio-demographic variables. Treatment cost for the last two

months and the responsibility of treatment cost coverage were also found to be statistically

significant in affecting the dependent variable with the expected signs.

The censored regression suffers from a number of specification problems. The disturbance

term is not normally distributed or homoskedastic. The fact that relaxing these two

assumptions of the model can make significant changes on the parameter estimates is

highlighted by the literature on the Tobit model. The results of the semiparametric

estimation technique shows that apart from family size, two other socio-demographic

variables, age and spouse deceased, were found to be statistically significant. However,

unlike the result of the parametric estimator, employment status was insignificant in

affecting WTP. The findings of this paper support the important role of income in

determining the WTP of respondents. Treatment cost and responsibility of coverage for this

cost were found to have the expected signs and were statistically significant.

In both estimation techniques, the initial bid price was shown to create bias on the final

response of respondents. However, this is one of the disadvantages of using a bidding game

as an elicitation method where even if individuals were randomly assigned five different

initial bid prices, the starting price had influenced the value respondents place on the drugs.

The study was, however, exempt from free riding problem and in general reliable.

Across the 25th, 50th and 75th quantile regressions, both income and family size are found to

be determining factors of the WTP of individuals, with expected signs. Comparing the

88
results of the 25th and 50th quantile regressions reveals that age, family size, income, total

treatment cost, responsibility of treatment cost payment and bid price are found to

significantly affect WTP for ARVs at a statistically acceptable level of significance.

However, when the results of the 75th and median regression are compared, only family size

and income are the common explanatory variables that are statistically significant.

Not surprisingly enough, the constant term is negative and found to be significant by all of

the estimators (Tobit, median and 75th quantile regression) at 1% level except by the 25th

where it is statistically significant at 5% level. This is simply a reflection of those 55%

respondents who were unwilling/unable to pay for the reason described earlier.

Since CLAD is an improvement over the Tobit model, only the results of the former

estimator are used for conclusion and policy recommendation.

5.4 Demand for ARV drugs

In this sub-section, an attempt is made to estimate the demand function of respondents for

ARV drugs.

Table 5.15 Estimated total WTP and demand

Respondents WTP at
WTP Interval ∧ Sample Total WTP least
WTP distribution that amount (cumm.)

(1) (2) (3) (4) (7)= (8) (9)

No % (2)*(3) No %
0-50 25 302 74.0 7,550 408 100
51-100 75.5 42 10.8 3,171 106 25.95
101-150 125.5 16 3.9 2,008 62 15.5

89
 151-200 175.5 5 1.2 878 46 11.25
201-250 225.5 3 0.7 677 41 10.05
251-300 275.5 5 1.2 1,378 38 9.35
301-350 325.5 6 1.5 1,953 33 8.15
351-400 375.5 6 1.5 2,253 27 6.65
401-450 425.5 11 2.7 4,681 21 5.15
451-500 475.5 1 0.25 476 10 2.45
501-550 525.5 3 0.7 1,577 9 2.2
551-600 575.5 1 0.25 576 6 1.5
601-650 625.5 1 0.25 626 5 1.25
801-850 825.5 1 0.25 826 4 1.0
901-950 925.5 1 0.25 926 3 0.75
951-1050 1000.5 1 0.25 1,001 2 0.5
1500-2000 1750 1 0.25 1,750 1 0.25
Total 408 100.0 32,302

In table 5.15 from the WTP intervals, WTP midpoints are calculated. The grand total WTP

(Birr 32,302) indicates that this is the amount all of the respondents living with HIV/AIDS

are expected to pay per month if ARV drugs are made available in the market.

In the table, it can be seen that as the monthly payment increases, the number of people

willing to pay the increased figures declines. From this relationship the demand curve for

ARVs can be derived by plotting WTP midpoints on the vertical axis and the number of

respondents willing to pay at least that class mark on the horizontal axis.

90
 Fig. 5.1 Estimated Demand Curve for
ARV Drugs
1800

1650

1500

1350
WTP (Midpoint)

1200

1050

900

750

600

450

300

150

0 50 100 150 200 250 300 350 400

No. of Respondents

The above demand curve has the form of a rectangular hyperbola, being asymptotic to both

axes, indicating a constant elasticity demand. This demand is usually associated with the

function (Frank, 1994) :

N = α / pη……………………………………………….(1)

Where: η and α are constant positive numbers.

N= PLWHA

P = Price (WTP midpoints)

α = some constant

η = elasticity coefficient

91
The implication of this function is, unlike the case of the straight line demand curve, here

people spend exactly the same amount when price is high as they do when price is low. As

one moves along the constant elasticity demand curve, the fall in the ratio of P/N is exactly

counterbalanced by the rise in the reciprocal of the slope.

It is useful to note that a demand curve of constant elasticity has its own particular algebraic

form. This form involves the logarithms of quantity (N) and price (WTP midpoints)

(Hirshleifer, 1980). Hence, (1) can be rewritten as:

Log N = log α - η log p......................................................(2)

It is the linear equation (2) that would actually be fitted statistically to the observation

reduced to logarithmic form. Therefore, estimating the double log function by fitting the

data indicates that α = 11.2. The implication of this result is respondents who spend their

entire allowance on ARVs each month would have a constant expenditure demand curve

for ARVs. The constant, 11.2, is equal to the amount of the allowance.

As mentioned above, η stands for elasticity. It follows that the proportionate change in the

number of people is a constant multiple of the proportionate change in price along the

demand curve. The estimation result for the elasticity coefficient is η = -1.448. According

to microeconomic theory, a constant elasticity demand curve with η > 1 has the property

that a price cut will always increase total expenditure and for one with η < 1 , by contrast, a

price cut will always reduce total expenditure (Frank, 1994). Hence, the coefficient of P,

92
1.448, indicates that whenever the price of ARVs declines, total expenditure of people

would increase.

Chapter six

CONCLUSION AND POLICY IMPLICATION

6.1 CONCLUSION

Ethiopia is one of the most seriously affected countries in the world by HIV/AIDS. The

country, with just 1% of the world’s population, contributes 9% of the worldwide cases of

HIV/AIDS. The country is also identified, among other four countries, to be hit by the Next

Wave of HIV/AIDS.

Clinical trials for vaccine against HIV are underway, which protect against certain viral

subtypes of HIV in the test countries. The vaccine probably will not be effective against the

most common subtypes of HIV in countries that will be hit by the Next Wave of

HIV/AIDS- Nigeria, Ethiopia, Russia, India and China leaving the countries to focus on

anti-AIDS education to bring behavioural change to prevent new cases and ARV drugs to

prolong the lifetime of people already living with the virus.

93
This study tried to answer the question: “How much are people living with HIV/AIDS

willing to pay for anti-retroviral drugs” and attempted to identify the determinants of

willingness to pay for the drugs. The contingent valuation method was employed to assess

the above two objectives. CVM can be a very effective preference elicitation tool as long as

respondents understand the context and content of survey instrument. The survey results

revealed that 99.5% of the sample respondents were aware of ARV drugs, though many had

hazy idea for how long the drugs extend life years and/or the side effects of the drugs.

With the exception of 32 individuals, generally, respondents showed interest in the drugs.

However, 48% of the respondents that showed interest claimed to have no ability to pay for

the drugs if the drugs are available in the market. Among those that stated different

amounts of willingness to pay, the mean WTP per month is Birr 163.64 per month if the

drugs are effective for indefinite years and Birr 148.48 if only for ten years. The median

WTP of the respondents if the triple combination lengthens life years indefinitely is Birr

100 per month and Birr 50 for ten years. The stated figures are far below the current market

price of the drugs in other parts of the world, Birr 250 per month (UNAIDS, 2002 and

Walgate, 2002).

In relation to income, individuals were found to be willing to spend from 25%-28% of their

income to the drugs per month.

To examine the effects of different variables, including socio-economic and demographic

characteristics of respondents, censored limited absolute deviation (CLAD) estimator was

94
used, whose result was compared to the Tobit results. CLAD is preferred because it makes

no assumptions about the nature of the WTP data. In other words, it does not impose

normality and homoskedasticity restrictions on the distribution of the error term.

The Tobit technique found employment status, income and treatment cost to significantly

and positively influence the WTP of individuals, whereas family size and the responsibility

of payment to the treatment cost were found to decrease the willingness to pay of

respondents significantly.

Results of the CLAD (50th quantile) estimation reveal that apart from family size, two other

socio-demographic variables, age and marital status (particularly those with spouse

deceased) were found to significantly determine the WTP for the drugs. Specifically, the

older the individual and the higher the income, the more would be the willingness to pay.

On other hand, those with spouse deceased, with larger family size and with others to pay

for their treatment cost have lesser willingness to pay for the drugs.

Finally, a comparison of the regression results for the 25th, 50th and 75th quantile regressions

indicated that some of the variables have different effects on the respondents’ WTP.

The attempt to derive the demand of respondents for ARVs resulted in a rectangular

hyperbolic curve implying a constant elasticity demand curve. The degree of

responsiveness of individuals for a 1% cut in price was found to be a 1.4% increase in the

demand for the drugs.

95
In general, the CV survey was found to be free of major biases and is reliable. It passed the

scope sensitivity and construct validity tests. Free riding was also not detected from the

responses of the individuals. However, even though the descriptive analysis showedd that

starting point bias did not exist in the survey, the contrary was found when bid price was

treated as an independent variable in the regression analysis. Both estimation techniques

found this variable to be highly significant.

6.2 LIMITATIONS OF THE STUDY

 As stated in third chapter of this paper, respondents were asked to assume two

cases: the drugs lengthening lifetime by indefinite and ten years. However, WTP

tends to be affected by the order in which the years are presented (Krupnick et al,

2000). This research therefore will not be exempted from this effect.

 The second limitation of the research is it attempted to derive the value of the drugs

by focusing exclusively on those who are in need rather than the general

community.

 The research was conducted in Addis Ababa and did not include people living with

the virus in the other parts of the country.

6.3 POLICY IMPLICATION

Among the different explanatory variables, income is found to be the most important

determinant of the willingness to pay for anti-retroviral drugs in both estimation techniques.

In the descriptive analysis, it was shown that 48% of the interviewed were constrained by

low income. This of course is the reflection of the general poverty of the country where it is
96
ranked 87 out of 88 developing countries in terms of human poverty index (UNDP, 2002).

This implies that the willingness to pay of the respondents for the drugs is highly

constrained by the low income. The inference from this would be the provision of anti-

retrovirals could not completely be left to the market. Different stakeholders- the Ethiopian

government, non-governmental organizations and PLWHA associations have to play their

role in the provision of the drugs at lower price. The fact that if prices are reduced people

would be more willing to pay is substantiated by the findings of the estimated demand

elasticity coefficient.

Treating opportunistic infections is costly- the average treatment cost for the two months
before the survey is Birr 255. For 63.2% respondents that went to a medical centre to be
treated the organizations they work in or the respective associations established by
PLWHA covered for the treatment expenses. One of the assumed positive impacts of the
availability of ARV drugs is the reduction in the frequent visits to medical centres for
the treatment of opportunistic diseases. Hence, it is possible for the associations to divert
their resource for the provision of the drugs to their members.
The survey result shows that 90% of those interviewed think the Ethiopian government
has not exerted its best effort to bring the drugs to the country. It can change this image
by offering special tax incentive or by providing subsidy to local investors to endorse
domestic generic production of the drugs. This may be expected to dramatically
decrease the price of ARV drugs that are supposed to be imported in the near future.

Almost 47% of the respondents who participated in the survey believe that non-
governmental organizations (NGOs) operating in the country should play a role in the
distribution of the ARV drugs. It is true that “improvement and advancement of the
quality of life of those who are disadvantaged and vulnerable” is included in the mission
statement of NGOs (Code of Conduct, 1999). Hence, NGOs operating in the country can
play a multifaceted advocacy role. The first is creating awareness among the community
of people living with the virus. The survey showed that more than 55% had no idea for
how long the drug extends life years or about the toxicity. Since NGOs work at grass-
root level, they could sensitize people on the issue.
NGOs could also involve themselves in constructive policy dialogue with the
government to undertake active measures towards the drugs.
The third advocacy role could be played with their parent organizations. Though NGOs
receive financial or material assistance from donors to operate their programmes,
donors’ interests are more in favour of prevention and control (UNAIDS, 2002).
International organizations functioning in the country, in particular, could convince their
97
parent organizations and donors to give attention to care and support to people living
with HIV/AIDS to raise funds for the importation of the drugs.

For further research, it is possible to include both people living with and without the
virus to elicit their willingness to pay for the drugs. This would enable researchers to
identify the value attached to the drugs by people falling into the two categories.
Furthermore, researchers should be able to avoid the limitations of this particular
research to check if results are different from what is indicated in this study.

REFERENCES:

Abelson, Peter. (1996). Project Appraisal and Valuation: General Principles and Six Case

Studies in Developing Countries.

Amemiya, Takeshi, (1985). Advanced Econometrics.

Bateman, J. and Turner, K. (1997). Valuation of the Environment, Methods and

Techniques: The Contingent Valuation Method, Environmental Economics and

Management.

Belaineh Taye (2001). Economic Cost of Malaria on Rural Households, and Willingness to

Pay for Bed Nets: The Case of Selected Rural Kebeles in ILU Woreda of West

Shoa Zone

Bishai, D.M and Lang, H.C (2000). The WTP for Wait Reduction: The Disutility of

Queues for Cataract Surgery in Canada, Denmark, and Spain, Journal of Health

Economics 19 (2): 219-230

Bonato, D., Nocera, S., and Telser,H (2001). The CVM in Health Care: An Economic

Evaluation of Alzheimer’s Disease

98
Cameron, A. (1988). A New Paradigm for Valuing Non-Marketed Goods Using

Referendum Data, Journal of Environmental Economics and Management,

15(3): 355-379

Chay, K and Powell, J., (2001). Semi Parametric Censored Regression Models, Journal of

Economic perspectives, vol. 15, #4

Cock, K. (2001). Choosing a Standard Regime, Porceedings of Consultative Technical

Meeting on ARV Therapy in Kenya.

“Code of Conduct for NGOs in Ethiopia”, 1999.

CRDA News (October 2002). “Together We Can Stop HIV/AIDS”

Diener, A., O’Brien, B., and Gafni, A. (1998). Health Care Contingent Valuation Studies:

A Review and Classification of the Literature, Health Economics, 7:313-326

ECA (Economic Commission for Africa). (January 2002). “HIV in sub-Saharan Africa”

FHI (Family Health International) (October 1998). “Impact on HIV” Vol. 1, number 1,

Floyd, K. and Gilks, C. (1997). Cost and Financing Aspects of Providing Anti-Retroviral

Therapy: A Background Paper.

Forsythe, Steven. (1998). The Affordability of Ant-Retroviral Therapy in Developing

Countries: What Policy Makers Need to Know.

Forsythe, Steven. (2001). An Economic Evaluation of HIV/AIDS Services In Kenya: A

Practical Policy Application of Contingent Valuation

Frank, R.H. (1994). Microeonomics and Behaviour, 2nd ed.

Freeman, M. (1998). The Benefits of Environmental Improvement: Theory and Practice

Gebregiorgis T. (2001). Antiretroviral Clinical and Management Issues, Proceedings of

Consultative Technical Meeting on ARV Therapy in Kenya.

99
Greene, W.H. (1993). Econometric Analysis, 2nd ed.

Greene, W.H. (2000). Econometric Analysis, 4th ed.

Gujarati, D. (1995). Basic Econometrics, 3rd ed.

Hirshleifer, Jack .(1980). Price Theory and Applications, 2nd ed.

“HIV 101: The Basic of Care” (2001).

Johansson, P.O. (1995). Evaluating Health Risks.

Johansson, P.O. (1996). Cost-Benefit Analysis of Environmental Change.

Johasson, M. et al (1997). Willingness to Pay for Reduced Incontinence Symptoms, British

Journal on Urology. 80(4), 557-562

Johnston, J. and DiNardo, J. (1997). Econometric Methods, 4th ed.

Jones, Charles. (1998). Introduction to Economic Growth.

Krupnick et al. (September, 2000). Age, Health and the Willingness to Pay for Mortality

Risk Reductions: A Contingent Valuation Survey of Ontario Residents.

McDonald, J.F. and Moffitt, R.A., (1980). The Issues of Tobit Analysis, Review of

Economics and Statistics, 62:318-321.

Ministry of Health. (2002) “AIDS In Ethiopia”. 4th ed.

Mitchell, R. and Carson, R. (1993). Using Surveys to Value Public Goods: The Contingent

Valuation Method.

Newman, Carol, (2001). A Double Hurdle Model of Irish Households’ Meat Expenditure

Patterns, Paper for the 71st EAAE Seminar, April 19-20.

NIC (National Intelligence Council) (2002). “Next Wave of HIV/AIDS: Nigeria, Ethiopia,

Russia, India and China” www.odci.gov/nic.

100
O’Conor, M., et al (1998). Urge Incontinence- Quality of life and patients’ Valuation of

Symptom Reduction ,pharmacoeconmics 14(5), 531-539.

Onjala, J.O. (June 2001). Multi-Nation Contingent Valuation, Environmental Economics

and Development.

Panos Institute (December, 2002). “Patents, Pills ,and Public Health: Can TRIPS deliver?”

Pol, V., and Cairns, J. (1999). Individual Time Preferences for Own Health: An Application

of a Dichotomous Choice Question with Follow-up, Applied Economics Letters

6(10): 649-654.

“Policy framework on the Provision and Use of ARV Drugs in Ethiopia” (March 2002).

Ethiopian Government.

“Policy on Antiretroviral Drugs: Supply and Use of the Federal Democratic Republic of

Ethiopia”. (2002).

Powell, J.L (1984). Least Absolute Deviations Estimation for the Censored Regression

Model, Journal of Econometrics, 25, 303-325.

Prescott, Nicholas (1998). Setting Priorities for Government Intervention with

Antiretroviral.

Sande and Mollering. (2000). Guide to HIV/AIDS Therapy, 9th ed.

Summary Guide (2002). “Economic Valuation with Stated Preference Techniques:

Summary Guide” Department of Transport, Local Government and Regions

U.K.

Tambour, M. and Zethraeus, N.(1998). Non-Parametric WTP Measures and Confidence

Statements, Medical Decision Making 18(3): 330-336.

101
UNAIDS (2002) “Assessment Report on Services Available in Addis Ababa to Diagnose,

Screen and Manage PLWHA, and Provide ARV Treatment of Opportunistic

Infections”.

UNAIDS (2002) “Care, Treatment and Support for PLWHA”

UNDP (2002. “Human Development Report: Deepening Democracy in a Fragmented

World “

UNFPA (2000). “State of World Population: People, Poverty and Possibilities”

Unger, R. et al (eds.) (2000) “Antiviral Against AIDS”.

Unicef/UNAIDS/WHO (May, 2001). “Sources and Prices of Selected Drugs and

Diagnostics for People Living with HIV/AIDS”

Walgate, Robert. (2002). As AIDS Drugs Prices Plummet for Third World, Questions Still

Around.

Walkowiak, Helena. (September, 1999). Improving Access to ARV Drugs in Developing

Countries: Constraints and Strategies.

Warner,M. (1996). International Encyclopaedia of Business and Management, Vol.3.

Whitehead et al. (May, 2000). Willingness to Pay for a Coastal Recreational Fishing

License: A Comparison of North Carolina Angler Groups.

Whittington, Dale (2002). Improving the Performance of Contingent Valuation Studies in

Developing Countries, (A paper prepared for special issue of Environmental and

Resource Economics)

Whittington et al. (2002). Private demand for a HIV/AIDS Vaccine: Evidence from

Guadalajara, Mexico.

102
WHO/UNAIDS (1998). Introducing Antiretroviral Treatments into Health Systems:

Economic Considerations, Guidance Modules on Antiretroviral Treatments.

Module 2.

Zweifel, p. and Breyer, F. (1997). Health Economics.

Table 5.1 socio demographic Characteristics of respondents

Variable Frequency Percent

Sex Female 264 60.0
Male 176 40.0
Marital Status Married 170 38.6
Divorced 84 19.1
Single 54 12.3
Widowed 132 30.0
Religion Orthodox 282 64.1
Protestant 82 18.6
Catholic 6 1.4
Muslim 65 14.8
Jehovah 5 1.1
Educational Status Illiterate 30 6.8
Able to read & write 20 4.5
Below 4th grade 28 6.4
5-8th grade 102 23.2
9-12th grade 196 44.5
Above 12th grade 64 14.5
103
Occupational Status Civil servant 33 7.5
Own business 76 17.3
Private employee 55 12.5
NGO employee 71 16.1
Unemployed 161 36.6
Daily labourer 31 7.0
Association worker 13 3.0

Table 5.2. Age, Family size and income of respondents

Variable Frequency Percent Min. Max. Mean
Age 18-25 80 18.2
26-33 167 38.0
34-41 134 30.5
42-49 45 10.2
50-57 12 2.7
18 65 32.79
58-65 2 .5
1-3 205 46.6
4-6 191 43.4
7-9 41 9.3 1 12 3.85
Family size 10-12 3 0.7
Income Below 120 144 32.7
150-250 78 17.7 25 5,000 395
300-400 77 17.5
450-550 50 11.4
600-700 33 7.5
750-800 11 2.5
Above 900 47 10.7

Table 5.3. Knowledge and practice of Respondents

Variable Yes No Do not know
No. % No. % No. %
Ever Heard 438 99.5 2 0.5 - -
Ever Used 28 6.4 412 93.6 - -
Know others who use 177 40.2 263 59.8 - -
Fair Price - - 438 99.5 2 0.5

Table 5.4. Attitude of Respondents towards ARVs

Variable Frequency Percent

Do not Know 243 55.2

104
 2-3 years 11 2.5
5-10 years 79 18.0
Effectiveness 10-15 years 44 10.0
15-20 years 34 7.7
20-25 years 8 1.8
25-30 years 3 0.7
Indefinite years 18 4.1
Toxicity Very high 28 6.4
Modest 66 15.0
Low 64 14.5
No toxicity 49 11.1
Do not Know 233 53.0

Table 5.5. Reasons for not paying

Reason Frequency Percent
Unable to pay 210 47.7
God 26 5.9
Unwilling to pay
For free 4 0.9
Side effect 2 0.5

Table 5.6. Willingness to pay amounts

WTP Frequency Percent
Interval Ten Indefinite Ten Indefinite
1-50 105 92 53.0 46.5
51-100 40 44 20.2 22.2
101-150 9 16 4.5 8.1
151-350 21 19 10.6 9.6
351-550 19 21 9.6 10.6
551-2000 4 6 2.0 3.0

Table 5.7. Summary statistics of WTP

Years Mean Median Min. Max.
Ten 148.4848 50 50 2000
Indefinite 163.6364 100 50 2000

Table 5.8. WTP and initial bid price

Willing to pay
Starting bid in Birr Yes No Total

105
 150 31 60 91
200 58 52 110
250 24 55 79
300 53 29 82
350 32 46 78
Total 198 242 440
Indefinite 10 years
Pearson 0.0024 0.044
Correlation
Sig. (2-tailed) 0.713 0.905

Table 5.9 WTP and Occupational status of respondents

Occupational 1-50 51-100 101-150 151-350 351-2000 Total
Status
Civil 10 6 1 4 3 24
Own Bus. 17 5 4 3 14 43
Private 9 10 4 2 1 26
NGO 17 13 4 10 7 51
Unemployed 27 6 2 0 2 37
Daily Lab. 6 0 0 0 0 6
Association 6 4 1 0 0 11

Table 5.10. Test for Free riding

Years Mean Median Min. Max.
Ten 75 50 50 350
Indefinite 80.56 50. 50 350

Table 5.11 construct validity test

Variable WTP indefinite WTP 10 Income
years years
WTP Pearson 1 0.968** 0.646**
indefinite correlation
years Sig. (2-tailed) . 0.000 0.000
WTP 10 years Pearson 0.968** 1 0.636**
correlation
Sig. (2-tailed 0.000 . 0.000
Income Pearson 0.646** 0.636** 1
correlation
Sig. (2-tailed 0.000 0.000 .
** Correlation is significant at the 0.01 level (2-tailed)

Fig 5.1 WTP and gender of respondents

106
 female

male
SEX

0
00
-2
00 50
15 -10
1 0
95 -95
1 0
90 -85
1 0
80 -65
1
60 600
1- 0
55 -55

107
1 0
50 -50

Fig 5.2 WTP and mean age of respondents

1 0
45 -45
1 0
40 -40
1 0
35 35
1- 0
30 -30
1
25 250
1- 0
20 -20
1 0
15 -15

WTP amount
1
10 00
-1
51
50
1-
60

50

40

30

20

10

0
Count
 0 0
00 00
-2 -2
00 0 00 0
15 105 15 105
1- 1-
95 950 95 950
1- 1-
90 850 90 850
1- 1-
80 650 80 650
1- 1-
60 600 60 600
1- 1-
55 550 55 550
1-

108
1-
50 500 50 500
1- 1-
45 450 45 450
1- 1-
40 00 40 400
4 1-
1- 35 350

Fig 5.3 WTP and mean Family size

35 50 1-
3
1- 30 300
30 00 1-
3
1- 25 250
25 250 1-
1- 20 200
20 00 1-
2
1- 15 150

WTP amount
15 50 1-

WTP amount
1 10 00
1-
10 00 -1
-1 51
51 50
50 1-
1-
50

45

40

35

30

25

20

1
Mean AGE Mean family size
Table 5.14 Results of the 25th and 75th quantile regressions

Variable 25th quantile regression 75th quantile regression

Coefficients t-ratio Coefficients t-ratio
Sex -0.1045171 -0.18 -0.2793702 -0.68
Age 0.0760673 1.97** 0.0369773 1.28
Marital status
married -0.9515558 -1.49 -0.5684495 -1.14
divorced -0.9156075 -1.1 -0.8509377 -1.13
widowed -0.8966771 -1.2 -0.6040108 -1.15
Religion
Orthodox/Catholic 0.3917815 0.77 -0.0069478 -0.02
Muslim -0.7984171 -0.88 -0.4257784 -0.85
Education
Primary 0.5679173 0.47 0.1384381 0.3
Secondary 0.7367327 0.65 0.129762 0.28
Post secondary 0.2115031 0.17 0.2730608 0.42
Employed 1.268851 0.79 0.6364031 1.73*
FS -0.3597291 -2.45** -0.3147552 -3.46***
FH -0.3764647 -0.74 0.8282734 1.79*
I 1.06995 4.1*** 1.830568 12***
YLV -0.0519969 -0.54 0.062187 0.88
HS -0.5004839 -0.6 -0.1733955 -0.32
TC 0.3621974 2** 0.1459139 0.88
WP -1.033206 -1.76* -0.4958644 -1.24
XY -1.599744 -1.3 0.1813678 0.21
BP 0.4741712 2.42** 0.1997491 1.63
FR 0.1815428 0.38 0.0982077 0.26
Cons -6.909576 -2.6** -4.935946 -4.55***

109
 Normality Test
Statistics WTP Income TC YLV
Number of Obs. 408 408 408 408
Skewness 3.313 .757 .990 1.280
Std. Error of Skewness .121 .121 .121 .121
Kurtosis 13.402 -.534 .240 1.185
Std. Error of Kurtosis .241 .241 .241 .241

WTP
400

300

200

100
Frequency

0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0

WTP

Multicollinearity test
Sex age married Divorced widowed Muslim Orth/
Cath

Sex 1.000
Age 0.396 1.000
married 0.104 -0.053 1.000
divorced -0.008 0.094 -0.292 1.000
widowed -0.119 0.223 -0.523 -0.244 1.000
Muslim 0.173 0.164 -0.021 0.078 0.007 1.000
OrthCath -0.075 -0.103 -0.034 0.124 -0.047 -0.170 1.000
primary -0.049 -0.087 -0.127 0.064 0.060 0.172 -0.017

110
secondary -0.072 -0.103 0.084 -0.056 -0.037 -0.155 0.059
Bsecondary 0.285 0.241 0.078 -0.007 -0.083 0.017 -0.033
employed 0.290 0.329 0.073 0.045 -0.028 0.100 0.002
FS 0.052 0.257 0.134 -0.057 0.022 0.078 -0.130
BP -0.002 -0.049 0.023 0.031 -0.027 -0.063 -0.022
YLV -0.005 0.234 0.014 -0.063 0.098 -0.009 -0.076
TC 0.030 0.168 0.001 0.073 0.080 0.062 -0.050
WP -0.125 -0.013 0.022 0.050 0.012 0.049 -0.014
FH 0.013 0.282 0.100 0.024 0.237 0.085 -0.089
Income 0.236 0.277 0.103 -0.007 -0.026 0.018 -0.019
XY 0.157 0.150 0.158 0.018 -0.112 0.048 0.010
HS -0.108 -0.010 -0.038 -0.021 0.081 -0.003 -0.082
FR -0.077 0.017 0.009 -0.076 0.062 -0.080 0.003

Primary Second. Bsecond. Employed FS BP YLV

Primary 1.000
Second. -0.593 1.000
Bsecond. -0.267 -0.390 1.000
Employed -0.142 0.050 0.231 1.000
FS 0.016 -0.006 0.017 0.075 1.000
BP -0.015 0.051 -0.063 0.043 -0.002 1.000
YLV -0.140 0.046 0.212 0.235 0.109 -0.012 1.000
TC 0.011 -0.053 0.092 0.109 0.061 -0.054 0.233
WP 0.076 -0.012 -0.130 -0.015 0.054 0.012 0.024
FH -0.077 0.031 0.022 0.222 0.225 0.116 0.137
Income -0.263 0.077 0.409 0.408 0.065 -0.003 0.319
XY -0.105 -0.018 0.231 0.150 0.026 0.036 0.045
HS 0.024 -0.027 0.018 -0.014 0.089 -0.042 0.122
FR -0.086 0.138 -0.039 0.049 0.050 0.097 0.110

TC WP FH Income XY HS FS
TC 1.000
WP 0.460 1.000
FH 0.150 0.065 1.000
Income 0.110 -0.155 0.138 1.000
XY 0.042 -0.066 0.131 0.257 1.000
HS 0.527 0.260 0.044 -0.093 -0.040 1.000
FS 0.043 0.079 0.107 0.049 0.010 0.006 1.000

111
 Test for outliers
30

292
20 104
86
185

301
217
99
268
282
10 199
2
294
288
299
150
269
163
103
309
305
182
33
76
202
408
221
205
112
113
364
144
124
405
59
407
29
390
389
404
290
277

-10
N= 408

WTP

QUESTIONNAIRE

Interviewer’s Name_____________
Date of the Interview____________
Supervisor’s Name______________
Date of Supervision_____________

Hello. My name is Martha Getachew. I am doing my Msc thesis on “willingness to pay for
antiretroviral drugs”, which I believe will help policy makers in their decision. The research
will be a successful story only if you help me by answering the questions below.
The interview is totally voluntary and if you feel uncomfortable in any of the questions you
are free not to answer.

PART I Socio-Demographic Characteristics

1. Sex_______

112
 2. Age_______

3. Marital Status
a. Married c. Single
b. Divorced d. Spouse deceased

4. Religion
a. Orthodox d. Muslim
b. Protestant e. Other___________
c. Catholic

5. Educational status
a. Illiterate d. Primary education complete
b. Able to read and write e. High school complete
c. Below 4th grade f. Beyond High school

6. Occupational Status
a. Civil servant e. Daily labourer
b. Own business f. Association employee
c. Private employee g. Unemployed
d. NOG worker h.Other___________

7. How many people live in your house (including yourself)?____________

PART II -Knowledge, attitude and practice

Studies show that using antiretroviral drugs lengthens the life years of people living with
the virus. These drugs are being distributed in the other parts of the world.

1. Have you ever heard/read/discussed about anti-retroviral drugs before this survey?
a) Yes
b) No
2. Do you use these drugs or have you ever used these before?
a) Yes
b) No
3. Do you know any one (family member or friend) who uses these drugs?
a) Yes
b) No
4. Do you think the current price in the country is reasonable?
a) Yes
b) No
c) Don’t know
5. For how long do you think these drugs lengthen the life of a person living with the
virus?_____________
113
 6. What do you think about the side effects of the drugs?
a. Very high toxicity d. No toxicity
b. Modest toxicity e. Don’t know
c. Low toxicity

PART III WILLINGNESS TO PAY FOR THE DRUG

Anti-retroviral drugs are assumed to lengthen the lifetime of people living with
HIV/AIDS for sometime and reduce the possibility of having opportunistic infections.
The Ethiopian government is thinking of providing ARVs for free only to prevent
mother-to-child transmission whereas other HIV carriers will be charged user fees.
This research will try to assess how much you are willing and able to pay for the triple
combinations.

Scenario I. Indefinite years

1. Let’s assume that triple combination lengthens the lifetime of people living with
the virus for indefinite years. Keeping in mind the limitations of your (family’s)
income, if the price of triple combination is Birr 200 per month are you willing and
able to pay for the drug?

a) Yes
b) No

Note to interviewer: If the response is “Yes” go to the next table; ask the figures till
the respondent refuses to pay. However, if the answer to the above question is
“No”go to quesion number 2

Birr per month Yes Specifically No

How much?
250-300
351-450
451-550
551-650
651-750
751-850
851-950
951-1050
1500-2000
2500-3000
3500-4000
4500-5000
Above 5000

114
 2. If the answer to question 1 is “No”, keeping in mind again the limitations of your
(or family’s) income, are you willing and able to pay from 150 per month for the
triple combination if the drug lengthens your lifetime by indefinite years?
a) Yes
b) No

Note to the interviewer: If the response is “Yes” to the above question go to next
table. However, if the answer is “No” go to the question number 3.
Birr per month Yes Specifically No
How much?
51-100
1-50

Note to the interviewer: If the respondents says “Yes” to any of the payment
questions go to question 4. However, if the respondent doesn’t show any interest in
the payment then go to the next question.

3. If you are not willing to pay at all for the drugs, would you tell me the reason?
Because:
a) I believe the drugs do not lengthen life years
b) I believe the drugs should be given for free
c) I have no income
d) I believe it has toxicity/side effects
e) I believe that only God can cure me
f) Other___________
4. If the drugs become available in the market and lengthen life years indefinitely
whose responsibility is to pay for the drugs for you?
a. Myself
b. My family
c. The organization I work in
d. The association I am a member in
e. Other___________________

Scenario II Ten years

5. Let’s assume that triple combination lengthens the lifetime of people living with the
virus for ten years. Keeping in mind the limitations of your (family’s) income, if the
price of triple combination is Birr 200 per month are you willing and able to pay for
the drug?

a) Yes
b) No

115
 Note to interviewer: If the response is “Yes” go to the next table; ask the figures till
the respondent refuses to pay. However, if the answer to the above question is
“No”go to quesion number 2

Birr per month Yes Specifically No

How much?
250-300
351-450
451-550
551-650
651-750
751-850
851-950
951-1050
1500-2000
2500-3000
3500-4000
4500-5000
Above 5000

6. If the answer to question 1 is “No”, keeping in mind again the limitations of your
(or family’s) income, are you willing and able to pay from 150 per month for the
triple combination if the drug lengthens your lifetime by ten years?
a) Yes
b) No

Note to the interviewer: If the response is “Yes” to the above question go to next
table. However, if the answer is “No” go to the question number 3.
Birr per month Yes Specifically No
How much?
51-100
1-50

Note to the interviewer: If the respondents says “Yes” to any of the payment
questions go to question 4. However, if the respondent doesn’t show any interest in
the payment then go to the next question.

7. If you are not willing to pay at all for the drugs, would you tell me the reason?
Because:

a) I believe the drugs do not lengthen life years

b) I believe the drugs should be given for free
c) I have no income
d) I believe it has toxicity/side effects
e) I believe that only God can cure me
116
 f) Other___________

8. If the drugs become available in the market and lengthen life years by ten years
whose responsibility is to pay for the drugs for you?
a. Myself
b. My family
c. The organization I work in
d. The association I am a member in
e. Other___________________

PART III- Determinants of Willingness to pay

1. How long has it been since you knew you have the virus in your blood?________
2. How do you evaluate your health status?
a. Very good d. Bad
b. Good e. Very bad
c. Modest
3. For the last two months, for how long did you visit a hospital, clinic or health station
for treatment?____________
4. For the last two months, without visiting a hospital, clinic or health station, did you
prescribe anti-pain for yourself?
a. Yes
b. No
5. For the last two months, how much did you spend for treatment cost?__________
6.If you incurred any treatment cost for the last two months, who covered for the expense?
a. Myself
b. My family
c. My organization
d. My association
e. Other___________
7. Are a family head?
a. Yes
b. No
8. If response to the above question is “Yes” how many persons depend on you?_____
9. Do you work and earn income?
a. Yes
b. No
If “Yes” how much do you earn?

Note to the interviewer: if the response is in terms of per month write the answer
under “monthly”- do the same for others accordingly.

Daily Weekly Monthly Annually

117
 10. Is there anyone in the house who works and earns income?
a. Yes
b. No
If “Yes” how much do they earn?
Number ofDaily Weekly Monthly Annually
workers
1.
2.
3.
4.

11. Do you get financial support from relatives, friends, association or others?
a. Yes
b. No
If “Yes” how much do you get?
Daily Weekly Monthly Annually

12. Is there anyone else in your family that is also HIV positive
a. Yes
b. No
Note to the interviewer: If the response to the above question is “Yes” ask the next
question; if “No” go to the next section.

13. What is your relation with this person?

a. Spouse
b. Child/children
c. Relative
d. Other___________
14. When the drugs become available in the market, are you expected to pay for the
person?
a. Yes
b. No

Part IV- General Attitude

1. In your opinion how do you think the distribution of the drugs be carried out?
a. Government should distribute for free
b. NGOs should distribute for free
c. Pharmacies should distribute the drugs by charging users
d. Other______________________

2. If your economic status changes for the better are willing to pay for the drugs?
a. Yes
118
 b. No

3. In your opinion do you think the Ethiopian government has exerted its best effort to
import the drugs to the country?
a. Yes
b. No

119