McMaster University AIDS pandemic Article Essay
Description
1. Summarize the article 2. Identify your causal question 3. What type of data does the article uses to tests its hypotheses(observational, social experiment, laboratory experiment)? 4. What estimation methods do the authors use? 5. What key results do the authors report? 6. internal strengths and weaknesses of these analyses? 7. external strengths and weakness of these analyses ?
1 attachmentsSlide 1 of 1attachment_1attachment_1.slider-slide > img { width: 100%; display: block; }
.slider-slide > img:focus { margin: auto; }
Unformatted Attachment Preview
THE HIV/AIDS PANDEMIC IN SOUTH AFRICA:
SECTORAL IMPACTS AND UNEMPLOYMENT
Channing Arndt
Assistant Professor
Department of Agricultural Economics
Purdue University
Jeffrey D. Lewis
Lead Economist, South Africa
The World Bank
May 2001
Conference on Global Trade Analysis
Forthcoming: Journal of International Development
The views expressed in this paper are those of the authors, and should not be
attributed to the organizations with which they are affiliated.
Abstract
South Africa is currently confronting an HIV/AIDS crisis. HIV prevalence in the
population is currently estimated at about 13% with that number projected to increase
over the next five years or so. Given the massive scale of the problem and the
concentration of effects on adults of prime working age, the pandemic is expected to
sharply influence a host of economic and non-economic variables. While the pandemic
will certainly influence the rate of economic growth, structural changes are also likely to
be one of the primary economic hallmarks of the AIDS pandemic.
This paper builds on the work of Arndt and Lewis (2000) who estimated the
aggregate macroeconomic impacts of the HIV/AIDS pandemic in South Africa using a
computable general equilibrium (CGE) approach. They found that, despite dramatically
lower rates of growth of the unskilled labor pool relative to the no AIDS trend,
estimated unemployment rates for unskilled labor in their base AIDS scenario
increased absolutely over most of the upcoming decade and are essentially the same
(slightly higher in fact) as the rates estimated for a fictional no AIDS scenario. In this
paper, we seek to further investigate the interactions between unemployment and AIDS
using the basic modeling approach set forth in Arndt and Lewis.
Before projecting the impacts of the pandemic on unemployment, recently
compiled historical data on employment, unemployment, and remuneration are presented.
The unemployment problem is, rather, an employment problem, and it is concentrated
primarily in the unskilled and semi-skilled labor category. Job creation performance over
the past three decades in this category has been dismal with total employment (formal
sector and informal sector) of unskilled and semi-skilled laborers in 1999 at only 92% of
the level present in 1970.
In a country with an extraordinarily complex historical legacy such as South
Africa, it is impossible to attribute this disastrous job creation performance to any single
factor. Nevertheless, large differences in remuneration trends across labor classes and
standard economic theory point to these trends as major contributing factors. By 1999,
real remuneration per unskilled and semi-skilled worker had grown to 250% of the 1970
level while remuneration for other categories had remained essentially flat. Based on
these data, the neoclassical conclusion that unskilled and semi-skilled labor has been
systematically pricing itself out of the market seems practically unavoidable.
Employment growth has, given slow economic growth rates, gone hand in hand with
wage moderation as in the highly skilled and skilled segments. In contrast, employment
compression has been associated with substantial real remuneration growth as in the
unskilled and semi-skilled segment.
With this historical background in mind, we turn to examining the interactions
between the AIDS pandemic and unemployment. Even though the pandemic is projected to
drive growth rates in the supply of unskilled and semi-skilled labor to around zero, our
analysis indicates that the pandemic will also depress labor demand leaving the
i
unemployment rate, in our base AIDS scenario, essentially unchanged compared to a
fictional no AIDS scenario. The pandemic depresses labor demand through three effects.
Declines in the rate of overall economic growth.
Pronounced declines in sectors that supply investment commodities, particularly
the Construction and Equipment sectors. These two sectors happen to use
unskilled and semi-skilled labor intensively and together account for a
significant share (16.3%) of total payments to this category of labor.
Beyond this investment demand effect (brought on by reduced savings), AIDS
induced morbidity effects on unskilled and semi-skilled workers tend to depress
output relatively more in sectors that use unskilled and semi-skilled labor
intensively with further negative implications for employment.
Countering these three effects will be key to palliating the negative economic
consequences of the pandemic and reducing unemployment rates. To reduce the
unemployment problem, South Africa must have rapid overall economic growth ideally
with sectors that use unskilled and semi-skilled labor intensively leading the way. Results
indicate that a policy of real wage moderation (or even modest decline) presents a
straightforward option for bolstering overall economic growth. A wage moderation
policy also provides a particularly large stimulus for sectors that use unskilled and semiskilled labor intensively with further positive implications for employment.
ii
THE HIV/AIDS PANDEMIC IN SOUTH AFRICA:
SECTORAL IMPACTS AND UNEMPLOYMENT
1. Introduction
South Africa is currently confronting an HIV/AIDS crisis. HIV prevalence in the
population is currently estimated at about 13% with that number projected to increase
over the next five years or so. The implications of the pandemic will be profound for
millions of families as the primary family wage earners and/or caretakers fall sick,
require care, and eventually die. The pandemic will, without doubt, place extraordinary
pressure on institutions that confront its direct effects, such as the health care system for
the care of those living with AIDS and social services/systems (broadly defined) for the
care of dependents of AIDS victims.
There is also fairly wide agreement that implications will not be confined to the
households and institutions in the direct path of the pandemic. This consensus stems
mainly from the massive scale of the problem. It is probably fair to say that any change
that reduces the population growth rate from around two percent to zero in the space of a
decade, as the pandemic is projected to do, could be expected to sharply influence a host
of economic and non-economic variables. Since the pandemic brings about this reduction
in population growth mainly by killing individuals of prime working age, there is
legitimate concern over economic impacts.
While the pandemic will certainly influence rates of economic growth, structural
changes are also likely to be one of the primary economic hallmarks of the AIDS
pandemic in South Africa. At a minimum, the onset of the pandemic implies major
departures from past trends in rates of accumulation of factors of production (e.g., skilled
and unskilled labor) and changes in consumption patterns of government and households
(e.g., more health care spending) with at least some of these consumption pattern changes
financed by switching from investment to current expenditure. These changes will
interact with policy and existing economic structure causing the economy to evolve
structurally in a manner that is likely to be quite different from the path in the absence of
the pandemic.
This paper seeks to build on the work of Arndt and Lewis (2001; henceforth A+L)
who estimated the aggregate macroeconomic impacts of the HIV/AIDS pandemic in
South Africa using a computable general equilibrium (CGE) approach. A+L sought
primarily to estimate impacts on gross domestic product (GDP) and absorption. However,
in their analysis, A+L also found that, despite dramatically lower rates of growth of the
unskilled labor pool relative to the no AIDS trend, estimated unemployment rates for
unskilled labor in their base AIDS scenario increased absolutely over most of the
upcoming decade and are essentially the same (slightly higher in fact) as the rates
estimated for a fictional no AIDS scenario.
Here, we seek to further investigate the interactions between unemployment and
AIDS using the basic modeling approach set forth in A+L. The paper is structured as
1
follows. Section 2 provides an historical perspective of the unemployment problem in
South Africa based on recently compiled data. Sections 3,4, and 5 draw heavily from
A+L in order to provide the context for a focus on the unemployment issue. In particular,
section 3 describes the modeling approach employed to simulate the economic
implications of the AIDS pandemic. Section 4 provides the primary assumptions
underlying the AIDS and no AIDS scenarios and section 5 presents basic
macroeconomic results. Section 6 analyzes the sectoral implications of the pandemic, the
implications of these changes in the structure of production for unemployment, and the
sensitivity of the overall outcome to the rate of growth of wages for unskilled and semiskilled labor. Section 7 concludes and discusses policy implications.
2. Unemployment in Historical Perspective
New data on unemployment rates by skill class are presented in Figure 1. 1 The
data show that, for all classes of labor, unemployment rates were quite low in the early
1970s. However, since 1976, unemployment rates for unskilled and semi-skilled labor
have increased essentially monotonically. In 1995, the unemployment rate, according to
this data, surpassed the mind-bogglingly high level of 50% and continued to climb for the
remaining four years of available data. In contrast, the unemployment rate for highly
skilled workers has been negligible throughout the period. The rate for skilled labor
began to climb more recently and has attained a fairly significant level.
The data in Figure 1 are based on a narrow definition of unemployment.
Broader definitions of unemployment, which are more generous in the definition of
actively and unsuccessfully seeking employment, would give even higher unemployment
rates while narrower definitions than the one employed would give lower rates. However,
given the unemployment rates obtained by any reasonable measure, debate over the true
magnitude of the unemployment rate is, for all practical purposes, moot. By any
definition, the unemployment rate among unskilled and semi-skilled workers is now
ridiculously high, and it has been increasing for nearly a quarter century.
FIGURE 1 ABOUT HERE
Figure 2 gives further insight into the unemployment problem in South Africa. It
is, rather, an employment problem. Job creation performance over the past three decades
in the unskilled and semi-skilled labor category has been dismal. Total employment
(formal sector and informal sector) of unskilled and semi-skilled laborers in 1999 was
only 92% of the level present in 1970. While the number of jobs in the informal sector
quadrupled between 1970 and 1999, the formal sector has been marked by massive job
shedding. Formal sector employment of unskilled and semi-skilled laborers in 1999 was
only three million compared with four million employed in 1970. The formal
employment peak occurred in 1981 at 4.4 million jobs. Between 1981 and 1999, the level
of formal sector employment declined in every year but three losing a total of 1.3 million
jobs in the space of less than two decades. The trend towards reductions in formal sector
1
The data series were derived from official South African statistical sources by Quantec Research.
2
employment of unskilled and semi-skilled labor is evidently firmly in place and shows no
sign of abating.
FIGURE 2 ABOUT HERE
In short, formal sector employment failed to grow as rapidly as the unskilled and
semi-skilled labor force in the 1970s and declined essentially throughout the 1980s and
1990s while labor supply continued to expand. In contrast, employment in the highly
skilled and skilled labor segments grew respectively by 380 and 200 percent over the
same period.
In a country with an extraordinarily complex historical legacy such as South
Africa, it is impossible to attribute this disastrous job creation performance to any single
factor. Nevertheless, large differences in remuneration trends across labor classes and
standard economic theory point to these trends as major contributing factors. Figure 3
shows real remuneration per employee by labor class since 1970. In 1999, real
remuneration per highly skilled worker was at 90% of the 1970 level while real
remuneration per skilled worker increased to 110% of the 1970 level. In contrast, real
remuneration per unskilled and semi-skilled worker in 1999 had grown to 250% of the
1970 level.
FIGURE 3 ABOUT HERE
Based on these data, the neoclassical conclusion that unskilled and semi-skilled
labor has been systematically pricing itself out of the market seems practically
unavoidable. Employment growth has, given slow economic growth rates, gone hand in
hand with wage moderation as in the highly skilled and skilled segments. In contrast,
employment compression has been associated with substantial real remuneration growth
as in the unskilled and semi-skilled segment.
With these employment and remuneration trends in mind, we turn to the task of
looking forward at employment trends in the context of the AIDS pandemic.
3. AIDS and Unemployment: An Economy-wide Approach
We estimate the interactions between the AIDS pandemic and unemployment using
a recursive dynamic computable general equilibrium (CGE) model of the South African
economy. CGE models have a number of features that make them suitable for examining
cross-cutting issues such as the impact of AIDS on key economic variables.
They simulate the functioning of a market economy, including markets for labor, capital,
and commodities, and provide a useful perspective on how changes in economic
conditions will likely be mediated through prices and markets.
Unlike many other partial equilibrium or aggregate macro approaches, they are based on
a consistent and balanced set of economywide accounts (called a Social Accounting
3
Matrix, or SAM), which requires (among other things) that key behavioral and
accounting constraints (such as budget constraints and balance of payments equilibrium)
are maintained, which in turn serves as an important check on the reasonability of the
outcomes.
Because they can be fairly disaggregate, CGE models can provide an economic
simulation laboratory with which we can examine how different factors and
channels of impact will affect the performance and structure of the economy, how
they will interact, and which are (quantitatively) the most important.
The model version employed here contains fourteen productive sectors, including
three service sectors of particular relevance to analysis of HIV/AIDS: medical and health
services, social services, and government services. There are five primary factors of
production (professional, skilled, and unskilled labor, informal labor, and physical capital),
five household categories representing income distribution quintiles, seven different
government functional spending categories, and three government investment
categories. 2
Sectoral production occurs according to a translog production function that
determines how capital and labor inputs are combined together in generating value added. 3
The value added aggregate is then combined with intermediate (material) inputs to produce
output according to a fixed coefficients technology. Profit-maximization by producers is
assumed, implying that each factor is demanded so that marginal revenue product equals
marginal cost. However, factors need not receive a uniform wage or rental (for capital)
across sectors; sectoral factor market distortions are imposed that fix the ratio of the
sectoral return to a factor relative to the economywide average return for that factor. The
productivity with which factors combine to produce output in each sector can be affected
in two ways: first, the overall productivity can be changed (corresponding to a change in
total factor productivity, or TFP) to reflect conditions in which the general productivity of
existing technologies is enhanced or reduced (i.e., with the same bundle of labor and
capital inputs, less output is produced), and second, the contribution of specific factor
inputs (such as labor) can be affected, implying that the effectiveness of each input unit is
reduced (i.e., even though there are 100 workers, the effective input is equivalent to only
95 workers).
2
The basic model data is derived from a 1997 Social Accounting Matrix estimated by WEFA, a South African
consulting firm. The WEFA SAM is in fact more disaggregated, including 45 productive sectors, and a
household structure differentiated by deciles (with the upper decile further broken down into five groups).
But for modeling purposes, we have chosen to use a more aggregated version of their original data.
3
The translog production function is a more flexible specification than the standard (non-nested) CES
production relationship often specified in such models, in that it allows for different elasticities of
substitution between input pairs (e.g. between capital and skilled labor, and capital and unskilled labor).
This is important in the South African context because of the enormous differences in factor utilization
(ranging from high unemployment among unskilled labor to full employment of professional labor) and
historic trends in factor use.
4
On the demand side, the South Africa model maintains the standard CGE
assumption that domestic goods are imperfect substitutes for traded goods (both exports
and imports). Sectoral exports are assumed different from output sold domestically, and are
combined using a constant elasticity of transformation (CET) function to form domestic
output. This treatment captures explicit differences between exports and domestic goods
(such as quality), as well as other barriers preventing costless reallocation of output
between the export and domestic markets (such as market penetration costs). Hence, the
price of the good on domestic markets need not equal the domestic price of exports, which
is determined by the world price, the exchange rate, and exogenous export subsidies.
Producers maximize revenue from selling to the two markets, so that the ratio of exports to
domestic sales is a function of the price ratio.
As with exports, sectoral imports and domestically produced goods are imperfect
substitutes in both intermediate and final uses. Demanders of imports minimize the cost of
acquiring a “composite” good, defined as a CES aggregation of imports and domestic
demand. 4 Substitution elasticities can vary by sector, with lower elasticities reflecting
greater differences between the domestic and imported good. Retaining the small country
assumption, the supply of imports is assumed infinitely elastic at a price fixed by world
market conditions. The domestic price of the imported good is determined by the world
price times the exchange rate, plus any tariffs. The assumption of cost minimizing behavior
by demanders implies that the sectoral desired ratio of imports to domestic goods is a
function of their price ratio.
Five household categories (based on income distribution quintiles) are distinguished
in the model, along with government and corporate accounts. Each group receives income
from a variety of sources and has explicit behavioral rules governing savings and
expenditure behavior. Firms receive the returns to capital, pay corporate taxes, and receive
transfers (from government and the rest of the world). The remainder is divided between
corporate savings and household dividends.
Factor income for each of the labor types is distributed to the households using
fixed shares, as are corporate dividends. Households are taxed at a fixed rate, and may also
receive transfers from the government; they save a fixed fraction of their income (which
adds to the pool of domestic savings), with the remaining income spent on consumption.
The government receives tax revenue from import tariffs, export taxes, indirect
(excise) taxes, the value added tax, income taxes on corporations and households, and the
proceeds from net foreign borrowing. The government spends money on transfers to
households and firms, seven different categories of goods and services (including separate
health and education), and three types of government investment. The remaining surplus or
deficit is added to the available supply of savings in the economy.
4
This characterization of imperfect substitutability was developed by Armington (1969). It has since become
a standard feature in numerous applied models; see, for example, Dervis, de Melo, and Robinson (1982)
and Devarajan, Lewis, and Robinson (1990).
5
Sectoral private consumption is determined through the fixed expenditure shares
under the assumption that households have a Cobb-Douglas utility function. Government
consumption is also allocated using fixed expenditure shares. Final demand for
intermediate goods is the sum of the intermediate demands generated in each producing
sector. Investment is allocated using dynamic updating rules discussed in more detail
below. Final demand for investment goods is obtained using an activity specific capital
coefficients matrix.
Several other features affect simulations with the CGE model. The savingsinvestment closure is savings-driven: in other words, the resources available for
investment each year are determined by the sum of savings generated by groups within the
economy (households, firms, and government) plus any foreign capital inflow. Government
current spending as a share of total absorption (C+I+G) is controlled exogenously. The rule
for allocation of government spending across the seven expenditure categories can
accommodate various crowding out mechanisms for example, increased spending on
health services can come entirely at the expense of other types of spending or through an
increase in the government deficit, which crowds out investment. In the base scenario,
AIDS-related government expenditure is deficit financed. Net foreign savings are fixed
exogenously, and the exchange rate varies to achieve external balance. In common with
other CGE models, the model only determines relative prices and the absolute price level
must be set exogenously. In our model, the aggregate producer price index is fixed,
defining the numeraire.
As a recursive dynamic model, the model contains a set of cumulation and updating
rules (e.g. investment adds to capital stock, after depreciation; labor force growth by skill
category; productivity growth). The purpose of these dynamic equations is to update
various parameters and variables from one year to the next, and for the most part, the
relationships are straightforward.5 Growth in the total supply of each labor skill category is
specified exogenously, and for the informal, unskilled and skilled labor groups (for which
inflexible wages leads to unemployment), the growth trajectory of real wages is also
provided. Sectoral capital stocks are adjusted each year based on investment, net of
depreciation, and investment is assumed to respond to differential sectoral profit rates so as
to preserve the rental rate differentials observed in the base year data. Sectoral productivity
growth (TFP) is specified exogenously.
Using these simple relationships to update key variables, we can generate a series of
growth scenarios, based on different assumptions for key exogenous variables (such as
demographics, government spending, and technology). Using actual data, we first run the
model forward from 1997 to 2000, providing an opportunity to calibrate the model to
recent actual performance (i.e., does the model adequately reproduce recent growth?). We
then use the model to generate forward projections from 2001-2010, based on the
assumptions and updating relationships provided.
5
Note that the model is not dynamic in the full economic sense of the word, since there are no multi-period
optimizing equations and hence no attempt to ensure intertemporal optimality. The simulations are best
thought of as a sequence of lurching equilibria, in which within period (static) equilibrium is first
attained, then the model lurches forward to the next period, and a new (static) equilibrium is found.
6
It is important to emphasize that, at present, we are primarily interested in the
differential impact of the HIV/AIDS pandemic. By using the model and varying our
assumptions about future trends in key variables (such as wage rates), we can use the CGE
model as a simulation laboratory with which to conduct controlled experiments. In
particular, we can compare a hypothetical no AIDS scenario to a series of more likely
AIDS scenario. From this vantage point, what matters most is whether our benchmark no
AIDS scenario is more or less reasonable, rather than whether it is accurate or not. In
other words, building more detail or feedback into the base no AIDS scenario will only
matter if we choose to vary these features across experiments: if they are not factors which
we believe will depend on AIDS-related variables, then including them should make little
difference to the differential among scenarios on which we are focusing.
4. Estimating The Macroeconomic Impact of the Pandemic
Box 1 summarizes the key channels incorporated into our base AIDS scenario,
along with the assumptions made in each case. An appendix provides details on
demographic assumptions.
Box 1: Features of the AIDS scenario
Effect
Population/labor supply: AIDS pandemic will
slow population growth and have differential
impact on growth in labor supply by skill category
Labor productivity: Incidence of HIV/AIDS
among workers will reduce labor productivity,
especially with onset of AIDS
Total factor productivity: Prevalence of
HIV/AIDS lowers overall productivity (due to
hiring and training adjustment costs, absenteeism,
slower technological adaptation, etc.)
Household spending patterns: HIV/AIDS
affected households will shift spending towards
health and related expenditures
Model Assumption
Slower growth in population and in labor force
by skill categories (taken from ING Barings,
2000; see appendix)
Effective labor input for each skill type reduced
proportionally with projected AIDS deaths
(from ING Barings, 2000) one period hence
Sectoral TFP growth declines with the onset of
the pandemic falling to one half the no AIDS
rate at the height of the pandemic
AIDS affected households save nothing and
increase their share of health services spending
to 10-15% of total spending (depending on
quintile), at the expense of other (non-food)
expenditures
Government spending: Spread of AIDS will Health share of total government recurrent
induce higher government spending on health and spending rises from 15% in 1997 to 26% in
social services, either displacing other spending or 2010; non-AIDS related spending remains a
constant share of total absorption.
increasing the deficit
Some more details on the AIDS scenario assumptions are worthwhile. With
respect to labor productivity, we assume that, due to morbidity and absenteeism, AIDSafflicted workers are half as productive as remaining workers. AIDS-afflicted workers
stay on the job for two years. The labor productivity effect for each skill class in period t
is assumed to be exactly proportional to the AIDS death rate for that skill class in period
t+1.
7
Reductions in TFP growth rates are keyed off of the unskilled labor AIDS death
rate. When that rate reaches its maximum value (3.4 deaths per hundred workers in
2010), TFP growth rates in all sectors are halved relative to the no AIDS scenario. The
AIDS TFP penalty in earlier years depends upon the ratio of the AIDS death rate for that
year to the maximum value. So, for example, in 2003, the AIDS death rate for unskilled
workers is 1.7 per 100 workers. The TFP growth rate in 2003 is two-thirds the no AIDS
growth rate [1.0/(1+1.7/3.4)=0.67]. 6 It must be emphasized that the TFP declines
simulated are, in large measure, hypothetical since very little solid information is
available on the implications of AIDS for overall factor productivity growth rates.
With respect to government spending, our assumptions on government
expenditure amount to a 6.9 percent annual real increase in government expenditure on
health from 1997 to 2010. This rate compares with a real rate of annual expenditure
increase (actual expenditure deflated by the CPI) on health recorded for consolidated
government accounts for the period 1992 to 1997 of 5.7 percent (South African Reserve
Bank, 1999). 7 Real expenditure increases for social programs (on orphans for example) is
assumed to be much lower at 2.7 percent per annum.
Since the focus is on unemployment, the growth trajectories of real wages are of
interest. Real remuneration for unskilled and semi-skilled labor is assumed to grow at a
2% annual rate, which is less than the approximately 3.5% average annual growth rate
(calculated as the simple average of annual changes) observed from 1970-1999. Real
remuneration for skilled labor is assumed to remain constant, which is a somewhat slower
growth rate than recent trend. Remuneration for highly skilled labor is set in the market
place.
5. Base Case Macroeconomic Results
Figure 4 provides a reference point for the basic characteristics of the no AIDS
scenario as well as some insight into the bottom line results on the economic implications of
the AIDS pandemic. The no AIDS scenario postulates relatively low average growth rates
during the 1997-1999 period (around 2.5 percent), consistent with South Africas
performance during this period. No-AIDS scenario growth rates accelerate slowly and
6
7
It may be worth highlighting the difference between the labor productivity effect and growth in TFP. The
labor productivity effect could be viewed as a stock effect. It says that the stock of labor at a point in time
is less productive as a function of AIDS death rates at that time. In contrast, the TFP effect could be viewed
as a flow effect. It says that the rate of technical progress is reduced as a result of the pandemic.
Even under a generous health care financing scenario, scarce government health resources will be
directed towards treating AIDS and related conditions at the expense of other uses. In other subSaharan African countries, AIDS expenditure appears to have severely crowded out of other health
expenditures, and there is some evidence of a decline in the health status for non-AIDS afflicted
populations (World Bank, 1997).
8
steadily throughout the simulation period, due primarily to capital deepening and projected
increases in the rate of accumulation of professional and skilled labor (which are projected
based on historical trends).
FIGURE 4 INSERT ABOUT HERE
The AIDS scenario offers a very different picture. Growth rates in the late 1990s
start off at roughly the same level as the no AIDS scenario, due to the (relatively) low
incidence of AIDS. However, the growth paths diverge significantly over the simulation
period, as the impact of the pandemic becomes more pronounced. GDP growth rates decline
from year to year through 2008 (to only around 1 percent), before rebounding slightly in
2009 and 2010. Differences in real GDP growth rates between the scenarios reach a
maximum of 2.6 percent in 2008.
These differences in growth rates cumulate over time to bring about a substantial
divergence in the overall size of the economy. Figure 5 shows that, by 2010, real GDP is
about 20 percent below the level attained in the no AIDS scenario. In comparing mediumterm performance, such real GDP measures are frequently used as an indicator of aggregate
economic welfare. With respect to the AIDS pandemic, there is the broad issue, which we
do not address, of whether any GDP or absorption-based indicator can provide an adequate
measure of welfare in the context of a pandemic that (among other effects) lowers average
life expectancy by around 20 years. We do attempt to address a blatant omission in a
traditional indicator. In particular, GDP in the AIDS scenario contains substantially
increased
