Title: Demand for health care services in Ethiopia: exploratory analysis based on household surveys
1Demand for health care services in Ethiopia
exploratory analysis based on household surveys
- Almayehu Geda
- Abebe Shimeles
2outline
- Background
- Objectives of the study
- Theoretical framework
- Estimation strategy
- Discussion of main results
- Policy implications
3Background
- Ethiopia spends close to 5 of GDP, equally
divided between public and private, on health
related services - This is not typical of Ethiopia, but, common
among low income countries in SSA (Figure 1)
4Figure1 Per capita expenditure on health as a
share of per capita consumption expenditure in
Africa (2005)
5But, health outcomes are not that encouraging
6Objectives of the paper
- What factors influence individual demand for
health care services that would improve health
outcome? - What are the lessons for public policy in
combating disease burden beyond the provision of
basic health care services?
7Theoretical framework
- Grossmans (1972, 2000, 2004) formulation of
individual health both as investment and
consumption good in a life cycle.
8The set up individuals maximize life time
utility subject to constraints
9Reduced form demand for health as investment good
10Standard demand function when health is treated
as a consumption good
11Data and key variable definition
- The paper uses the 2004/06 Household Income
Expenditure Survey that covered around 21,000
households close to 100,000 individual stories - Dependent variable (health outcomes)
- A dummy if an individual has been sick in the
last four weeks - Number of days lost due to illness
- Stunting
12Variable definition
- Health inputs and control variables
- Affordability, distance to nearest health post,
health center clinic, quality of services,
perceptions about health care systems - Highest grade attained
- Real per capita consumption expenditure, calorie
intake and employment status - Demographic characteristics (age, household size,
gender) and regional dummies
13Empirical strategy
- In the health demand model, key individual
characteristics may potentially be endogenous - Particularly education attained and per capita
income level are likely to be correlated with the
error term for a number of reasons
14Empirical strategy
- We use two approaches to deal with these
problems, particularly with the education
variable. - First, estimate structural model (joint
determination of health demand and educational
attainment in a two-step procedure) - Second, using instrumental variables (distance to
nearest primary and secondary schools) - Both approaches led to valid orthogonality
condition between regressors and error terms.
15Descriptive statistics
- Large family size (5.6), young population (mean
32 years) and largely uneducated labor force. - The highest grade completed in the country is
grade 11 with 46 having never been to any formal
school in their life. Poverty is the main reason
for failing to go to school.
16Descriptive, contd..
- Close to 60 of individuals attributed lack of
money as the main reason for not going to school.
- About 7 also said that they had to work instead
of going to school probably to support family
businesses. - Access explains only 16 of the reason for
failing to go to school.
17Descriptive..
- Family formation explains a significant
proportion (10) of avoiding school probably more
for women particularly in rural areas. Other
reasons include disability and sickness which
affect school attendance. - Negative attitude is also one of the important
variable explaining aversions towards schooling
(10)
18Key results on demand for health as investment
good
- Dummy if the individual is down with illness in
the last four weeks (Table 3 Table 4) - In rural areas, the effect of individual
demographic characteristics (age, sex, marital
status, household size) depend on the choice of
estimation method. - In urban areas, being a female and older has a
high risk of falling sick despite estimation
methods.
19Key results ..
- Larger family size in both areas imply less
vulnerability to illness episodes, but up to a
point. Very large families tend to suffer from
health shocks (combination of initial health
endowment and reduced nutrition).
20Key results
- Education plays an important role in rural areas
in enhancing the efficiency of individual level
health production. This is not the case in urban
areas, possibly due to the weakness of distance
as an instrument for education.
21Key results (Table 5)
- Number of days lost due to illness is clearly and
significantly affected by the level of education
attained by the individual with elasticity of
about 0.5 for the whole sample and 0.2 for rural
areas. - Per capita consumption is also very important in
the whole sample as well as in rural areas. - Both variables are not important in urban areas.
22Key results Table 5
- Health inputs (affordability, quality, distance,
attitude to modern health care) seem to have
large and significant impact in urban areas than
rural areas. - Employment status influences health outcomes in
both rural and urban areas
23Key results table 6
- Education, particularly early access to formal
school, has a large and significant impact on
stunting
24Demand for health as a consumption good
- Estimates based on Linear expenditure system
indicates that health demand is inelastic with
respect to income (0.6) and price elastic as well
(0.4) which is not surprising. - Health is a necessity commodity (Figure 2)
25Figure 2 concentration curve on health and other
commodities
26Public policy implications
- The fact that education plays a key role in
affecting health outcomes by enhancing efficiency
implies that investment in education has a direct
and large impact on health. - Coordinating health policy with education policy
could reduce enormously health burden faster and
cheaper.
27Public Policy implications
- Distance to nearest health care services matters,
but not so much as cost, attitude, and quality of
the services (Annex Table 1) - Regional distribution of health care services
seem not to be that important which is
interesting, perhaps indicating that access to
heath care may not be region specific
28Thank you!