Does the Private Sector Care About AIDS?
Evidence from Firm Surveys in East Africa
Manju K. Shah
Ginger L. Turner
Keywords: Economics, East Africa, AIDS, private sector
Objective: Our objective is to identify the determinants of HIV/AIDS prevention
activity and pre-employment health checks by private firms in Kenya, Uganda and
Design: We use data from the World Bank Enterprise Surveys for Uganda, Kenya and
Tanzania, encompassing 860 formally registered firms in the manufacturing sector.
Methods: Econometric analysis of firm survey data is used to identify the determinants
of HIV/AIDS prevention including condom distribution and voluntary counseling and
testing (VCT). Multivariate regression analysis is the main tool used to determine
Results: About a third of enterprises invest in HIV/AIDS prevention. Prevention activity
increases with size, most likely because larger firms, and firms with higher skilled
workers have higher replacement costs. But even in the category of larger firms, less
than 50 percent provide voluntary counseling and testing (VCT). We find that the
propensity of firms to carry out pre-employment health checks of workers also varies by
size of firm and skill level of the workforce. Finally, data from worker surveys show a
high degree of willingness on the part of workers to be tested for HIV in the three East
HIV/AIDS has had an enormous impact on the economies of sub-Saharan Africa. Using
firm survey data from East Africa, we find that despite the high sero-prevalence rate of
HIV in the survey area, only a small proportion of firms—about 35 percent--engage in
prevention activities. And while prevention activity increases with firm size, less than 50
percent of large firms provide voluntary counseling and testing (VCT).
Using data from the World Bank Enterprise Surveys, which includes a sample of 860
firms in Uganda, Kenya and Tanzania, we examine two discrete actions--providing
prevention services, and conducting pre-employment health checks. We find that about
35 percent of firms engage in HIV/AIDS prevention activity, while the percentage of
firms conducting pre-employment health checks in our sample ranges from about 20
percent in Uganda to over 50 percent in Tanzania. Finally, our data also indicate that a
large proportion of workers are willing to pay to be tested for HIV/AIDS.
We find that larger firms and firms with a higher skilled and/or better trained workforce
tend to do more about HIV/AIDS through prevention activities; these firms are also more
likely to conduct pre-employment health checks to screen applicants. Firms where a
majority of workers are unionized are also more likely to carry out HIV/AIDS prevention
activities and pre-employment health checks. Finally, managers who are concerned
about absenteeism are also more likely to carry out HIV/AIDS prevention activities. We
also find that workers have a high willingness to be tested for HIV when asked; this is not
consistent with available private sector data on VCT uptake and therefore suggests that
significant barriers remain for workplace provision of VCT.
II: Economic Analysis of HIV/AIDS in East Africa—Review of the Literature
Kenya, Tanzania and Uganda have all been struggling with the problem of HIV/AIDS for
more than a quarter-century. Table 1 below presents the HIV prevalence rates for these
countries in 2003 (the period of data collection used in this paper). We see that Tanzania
had the highest prevalence rate and absolute number of HIV-positive persons, followed
by Kenya, and Uganda. All three countries have mounted public campaigns to fight
HIV/AIDS; these campaigns have increasingly been supplemented by private sector
While there is a large literature on the problem of HIV/AIDS in Africa, there is relatively
little rigorous analysis of private sector activity. A global survey in 2003 revealed that the
private sector is not doing enough about AIDS (Bloom, 2004; Taylor et al, 2004). The
World Economic Forum’s Global Health Initiative website summarizes the results of the
study as follows:1
Of the nearly 8,000 businesses surveyed in 103 countries:
-47% felt that HIV will have some impact on their business; this number is much
lower in countries that to date have not been hard-hit by HIV. There are
important regional variances – in Africa, 89% thought HIV would have some
impact, but in the Middle East and North Africa that figure dropped to 33%.
Worldwide, 21% of surveyed firms feel that HIV will have a severe impact on
-Business leaders estimate lower HIV infection rates among their workforce than
UNAIDS (official national adult prevalence figures), although 36% of business
leaders did not or could not estimate how many of their employees had HIV. The
small proportion of firms that have conducted quantitative studies estimates lower
rates than other firms.
In summarizing the findings of their paper, David Bloom and coauthors argue that firms
have taken little action regarding HIV in Africa (Bloom et al, 2004). They write that the
largest discrepancy between firm perceptions and actual data is to be found in Africa,
where 45 percent of firms report less than 1 percent prevalence, despite data from
UNAIDS that shows only 10 percent of respondent firms in Africa are located in low-
prevalence countries. They argue that as of 2003-04, the response to AIDS by the private
sector has been piecemeal with only a few firms having HIV/AIDS policies; the response
is limited even when firms are quite concerned about HIV. In these cases, businesses
seem to rely more on the public sector to deal with the problem.
In Rosen’s analysis of Nigeria, she also argues that managers are doing little about AIDS
(Rosen, 2001). Survey data used in this paper in 2001 showed that AIDS was not yet a
big problem in the Nigerian workplace and most managers have had little experience
dealing with it. Rosen also makes the interesting argument that given the high cost of the
business environment in Nigeria (power, water), it is unlikely that AIDS would enter the
“top ten” list of concerns for a while. Using a similar firm-level dataset, Biggs and Shah
looked at the impact of AIDS in the mid-1990s through worker attrition due to sickness
and death on firm performance and concluded that there is no significant measurable
impact (Biggs and Shah, 1997).
A recent study of agricultural workers in Kenya provided empirical estimates of the
impact of HIV/AIDS on labor productivity, by comparing healthy workers to workers
who later left the company due to HIV, through retrospective measures of output for
several years before their exit (Fox et al, 2004). Workers terminated because of AIDS-
related causes earned 16-18 percent less in the two years before termination, as well as
choosing less strenuous tasks and using more sick leave days. Rosen et al. 2004
examined the cost of AIDS to six large employers in South Africa, estimating the cost at
0.4 to 5.9 percent of the total wage and salary bill, with each infected employee costing
the employer an average of 0.5 to 3.6 times his or her annual salary. Rosen observes
elsewhere that many large employers are actively taking steps to shift the economic
burden of AIDS onto employees and governments, through such practices as outsourcing
unskilled jobs and capping benefits premiums (Rosen and Simon, 2003).
In a survey of 80 small and medium enterprises in South Africa, Connelly and Rosen
(2005) found that managers on average ranked HIV/AIDS as 9 out of 10 on the list of
priorities. Managers attributed a low percentage of productivity losses to HIV and found
worker replacement inexpensive. In addition, the study found lack of information to be a
major constraint; managers were unaware of free services available nearby. Aurum
Health recently demonstrated the profitability of AIDS workplace programs in 9 large
firms, including Anglo-American Mining. It observed a 60 percent decrease in
absenteeism, which compensated for 70 percent of the costs of the AIDS workplace
programs, the rest of which were covered by other cost savings (Aurum Health, 2005).
The South African Business Coalition Against HIV/AIDS (SABCOHA) has recently
targeted SMEs with its SME toolkit, for sale for approximately $215, which has attracted
very low uptake (Mears, 2005).
A number of studies have quantified the projected macroeconomic impact of HIV/AIDS
on the labor force (Over, 1997; Dorrington, 2002; Coulibaly/ILO, 2005). It has been
more challenging to demonstrate the microeconomic cost to firms of HIV-related
absenteeism and lower productivity, mostly due to the difficulty of gathering firm data
and confidential worker health information (often not known by managers or even
workers themselves) within the same survey instrument. Although Biggs and Shah found
no significant impact of HIV on productivity across a large survey of manufacturing
firms, possibly due to the ease of replacing workers in the earlier years of the epidemic,
more recent case studies (Fox et al, 2004; Maffessanti, 2006a and 2006b) have been able
to identify a link between HIV infection, higher absenteeism, and lower productivity.
III. Firm Behavior in East Africa
The analysis contained in this paper is based on a sample of 860 firms across Uganda,
Kenya and Tanzania and 4,955 workers. The data for this study come from the World
Bank’s Enterprise Surveys, collected in 2002-03, in collaboration with local
organizations in Africa.2 These firms are located in “traditional” sectors such as agro-
processing, wood/furniture, textiles/garments/leather, paper and publishing, construction,
chemicals and plastics, and metalworking. Each firm was interviewed in person by a
team of enumerators--in most cases, the manager, accountant, and up to 10 workers were
interviewed separately to collect the information used in this analysis.
2 More information about this dataset is available at www.enterprisesurveys.org. The collaborating
institutions for the design and enumeration of the East African surveys were the Kenya Institute for Public
Policy Research (KIPPRA), the Economic and Social Research Foundation-Tanzania (ESRF) and the
Uganda Manufacturers Association Consulting Services (UMACIS).
The sampling strategy that was followed was standardized across the East African
surveys. In each of the three countries, a stratified random sample was drawn, based on
available data regarding the population of formal, registered firms.3 Firms were sampled
primarily by number of employees and secondarily by sector and region. In Kenya, a
sample of 284 firms were surveyed, constituting a sampling rate of 15.2 per cent for
formal manufacturing firms. About 62 per cent of the firms were located in the Nairobi
and 10.6 per cent in the Eldoret/Kisumu region. Other key regions included in the survey
were Mombasa and Nakuru.
In Tanzania, the survey covered 276 firms from the manufacturing sector. These
manufacturing firms operated in 8 industrial sectors and in 10 regions of mainland
Tanzania, as well as Zanzibar. The sectors and regions covered in the survey were
selected because they had relatively high concentrations of manufacturing firms. The
regions covered in the survey were four regions on the eastern coast (Arusha,
Kilimanjaro, Tanga, and Dar es Salaam), three in the center (Morogoro, Iringa, and
Mbeya), one island (Zanzibar), and three in the north (Kagera, Mwanza, and Mara).
Most of the manufacturing firms in the survey were small, with the median of 31
employees. Larger firms were most common in Iringa/Mbeya and Tanga, where the
median establishments had 234 and 68 workers respectively.
Finally, in Uganda, the sample was drawn following the same rules as Tanzania and
Kenya, stratified by size, and then sector and location. With assistance from the Uganda
Bureau of Statistics, a sample of 300 manufacturing firms were surveyed, covering the
Central, Northeast and Southwest regions and the major areas of manufacturing activity.
It is important to note that in each survey, firms with a larger number of employees were
correspondingly more likely to be drawn--this method of sampling enables an adequate
number of firms to be sampled in each size class. Consequently, the observations are
3 It is important to note that these surveys do not cover firms in the informal sector.
weighted, where relevant, according to the probability of being sampled.4 Appendix 1
shows the distribution of firms in each of the three surveys, by size, sector, and location.
In the survey data, we identify two actions that firms may take in response to the
HIV/AIDS impact on the workplace:
1) Conducting prevention activities
2) Conducting pre-employment health checks of workers
HIV/AIDS prevention activities are defined as the following activities in the survey:
Prevention messages, which mostly consists of putting up posters around the
Distributing condoms on the premises of the firm
Providing HIV/AIDS counseling
Offering Voluntary Counseling and Testing (VCT)
Before discussing our results, it is worth asking why firms invest in prevention activity at
all. One might argue that the beneficial effects of prevention activities are low because of
high turnover rates or because of the length of time before people become ill with HIV.
Similar questions are often asked about why firms invest in training workers. We
recognize the validity of the question and offer several plausible explanations. First, the
average length of tenure for a full-time worker is fairly long--9 years in Kenya, 7 years
in Tanzania, and about 5 years in Uganda. Second, the length of time before people
become ill with HIV is not very long in East Africa—it could be as short as 2 or 3 years.5
Third, firms may be trying to retain atleast some types of workers but cannot target them
and therefore carry out prevention for the entire workforce. Firms may consider these
activities to be an extra benefit to the workers, may use them as signalling devices to
4 Detailed tables of the sample for each country are available from the authors and can also be found in the
World Bank’s Investment Climate Assessement for each country, available at www.worldbank.org/rped.
5 Atleast two studies have shown that the length of time before the onset of illness is quite short in Africa
(N’Galy et al 1988; Whittle et al 1992). However, one study reported a longer time from serocoversion to
AIDS (Morgan et al, 2002).
attract and retain better or more skilled workers, and may see gains to worker
productivity from these investments. Fourth, the activities we describe in this paper—
posters, condom distribution, and VCT--are generally not very costly to implement.
Perhaps most importantly, prevention activities may be driven by what the manager
believes to be its returns, based on the manager’s beliefs about the nature of HIV.6
We find that overall, about 35 percent of all firms in our sample conducted HIV
prevention activities. Of this set of firms, 15.6 percent of firms provided HIV education
(prevention messages via posters) and distributed condoms, while another 19.5 percent
conducted these activities as well as VCT. Table 2 shows HIV/AIDS prevention activity
by country and by region.
Prevention activity in Tanzania is lowest, which is interesting, given that Tanzania has
the highest HIV prevalence rate. Uganda has the highest proportion of firms putting up
prevention messages and distributing condoms. These activities tend to be less
expensive, and may be driven by increased awareness created by publicly-funded
programs. Kenya has the highest percentage of firms engaging in counseling and testing.
Both Kenya and Uganda have visible public-awareness campaigns to fight HIV/AIDS—
this may have some effect in terms of influencing firms to undertake prevention
Table 2 also shows that prevention activity has a positive association to regional HIV
prevalence.7 Firms in high-prevalence regions are almost twice as likely to engage in
prevention activity than firms in low-prevalence zones, for each category of prevention.
High-prevalence regions such as Nyanza in Kenya; Iringa, Mbeya, and Dar es Salaam in
Tanzania; and Kampala and Entebbe in Uganda, are more likely to have firms that
6 We do not have information about antiretoviral programs in this dataset. However, in the period 2002-03,
it is highly unlikely that workers in any firm in East Africa had access to employer-provided ARV
treatment. These treatments were prohibitively expensive and restricted to a few individuals, at best. Only
South Africa was beginning to see ARV provision in this time period.
7 Low, medium and high prevalence are defined as the following: Low is <5% population infected with
HIV, medium is 5-10%, high is >10%
conduct prevention activities. It is however worth pointing out that the vast majority of
our firms (60 percent) are in medium or high prevalence zones.
There are only small differences in prevention activity across sectors—the
agroprocessing/food sector has the highest share, followed by the furniture/wood sector
and the construction/machinery sector, and prevention activity is least likely in the
textile/garments/leather sector. Prevention activity varies in about a 10 percent range and
there is no obvious difference between these sectors with respect to labor intensity. Table
3 reveals interesting differences in prevention activity by firm size. Not only are large
firms more likely to do prevention activity, but they are more likely to do more high-cost
prevention activity, such as voluntary counseling and testing (VCT), and financial aid for
employees. The figure is highest for large and very large firms in our sample; close to 50
percent of large firms and over 70 percent of very large firms are engaged in some type
of prevention. Overall, about 35 percent of firms carry out some sort of prevention
Firms which train workers are twice as likely to engage in prevention activity and 60
percent of firms that do AIDS prevention also provide training to their workers.
Similarly, 61 percent of firms that provide VCT also provide worker training; only half
that percentage provide training in the category of firms that do not provide VCT and
other high-cost services.
Finally, do firms do more prevention activity when the perception of worker-absenteeism
is higher? Our data show that firms reporting a higher rate of absenteeism are more
likely to conduct HIV/AIDS prevention activities. About 43 percent of firms that say
that absenteeism is a problem carry out HIV/AIDS prevention activities; this number falls
to 29 percent for firms that do not report absenteeism as a problem. One explanation is
that managers who are likely to observe absenteeism may also be more likely to do
8 The Tanzania survey asks about “high” HIV-related and general absenteeism ; the Uganda and Kenya
surveys ask about “high” HIV-related absenteeism and “increased” general absenteeism. The question
does not specify the time period of increase but the last 12 months is clearly implied from the flow of
Pre-Employment Health Checks
About a third of firms in our survey engage in pre-employment health checks. Pre-
employment health checks that do not specifically test for HIV/AIDS may not detect
workers’ HIV infection status. Our survey does not ask whether pre-employment health
checks include HIV testing, or whether managers understand that HIV/AIDS status
would be visible from general health examinations. Some managers may not make the
connection between HIV prevention and general health testing; others may make guesses
as to the reasons for symptoms observed during the pre-employment check.
Pre-employment health checks of workers are controversial, to say the least. Opinions
vary about whether pre-employment health checks are illegal in East Africa. There are
national policies in existence that ban HIV testing of workers in Tanzania and Kenya, but
these do not seem to be implemented in a uniform manner. While it appears that HIV
testing as a condition of employment is illegal, the law appears to be less clear on the
issue of pre-employment health checks. These checks are largely conducted outside the
employer-employee relationship i.e. prior to the potential employee being hired. It is also
unclear to what extent pre-employment health checks can identify the HIV status of the
employee; it may well be the case that potential employers are making no more than a
guess about HIV status, particularly in cases where the CD4 count can be ascertained.
Table 4 shows the incidence of pre-employment health checks, by country and by region.
About 33 percent of firms in our sample engage in pre-employment health checks of
The proportion of firms conducting a pre-employment health check of workers is highest
in Tanzania (51.9 percent), followed by Kenya (34.5 percent) and Uganda (19.7 percent).
Uganda and Kenya do more HIV/AIDS prevention compared to Tanzania, in our sample
of firms. Tanzania also has the highest HIV/AIDS prevalence across the three countries,
as reported in the first section of this paper. Country-wide prevalence rates could
influence firms’ concerns about HIV/AIDS and therefore cause them to conduct health
checks in the hiring process. We also see in Table 7 that pre-employment health checks
are positively correlated with the prevalence of HIV.
Like prevention activity, pre-employment health checks vary by firm characteristics.
They are more likely to be carried out by foreign-owned firms; our data show that almost
50 percent of foreign-owned firms carry out health checks versus 30 percent of
domestically-owned firms. Table 5 shows the incidence of pre-employment health
checks by firm size. The proportion of firms performing a pre-employment health check
of workers increases with size, with over half of large and very large firms engaging in
The proportion of firms performing a pre-employment health check is highest in the
agro/food processing sector (about 50 percent), followed by the chemicals/plastics and
textiles/garments, and is lowest in the paper/printing sector. There is no obvious reason
for this difference among sectors, but it may be caused by a third factor, such as the
differences in worker demographics and education levels across sectors. Health concerns
may also be higher in the food industry, for safety reasons.
Pre-employment health checks are likely to be higher for firms that invest in worker
training and worker replacement is costly, and for firms with a higher skill composition
of their workforce. Our data show that 56 percent of the firms that provide pre-
employment health checks also provide training to their employees; this number drops to
34 percent for firms where no pre-employment health check is carried out. Firms that do
pre-employment checks also have a slightly higher ratio of skilled workers than those that
do not—38 vs. 34 percent.
III. Econometric Estimations of Firm Behavior
In this section, we examine the determinants of HIV/AIDS prevention activities and pre-
employment health checks, in a multivariate framework, using a Probit model. We base
our econometric anlaysis on a simple cost-benefit model, as discsused earlier.
This model leads to several hypotheses:
(1) Firms which use a higher ratio of skilled labor are more likely to invest more in
AIDS prevention and/or pre-employment health checks because of higher replacement
(2) Firms which carry out training programs are more likely invest more in HIV/AIDS
prevention and/or pre-employment health checks because of a higher level of investment
in employees. It is important to note that skill ratios are independent of whether the firm
invests in training of workers; skill ratio is defined by job status i.e. the ratio of managers
and professionals to total workers. In each skill category, the firm may or may not
provide formal training. Therefore, the first hypothesis captures formal schooling (pre-
employment human capital formation) while the second captures post-employment
(3) Prevention activity varies across sectors according to the degree of mobility in each;
firms will invest more in HIV/AIDS prevention in sectors where workers are less mobile.
We control for firm-specific characteristics such as size, ownership and degree of
unionization as a measure of bargaining power of labor. The hypothesis is that a more
unionized labor force will lead to greater HIV/AIDS prevention activity. Related to this,
it may also lead to more pre-employment health checks as firms anticipate that they need
to offer a higher level of services to their unionized employees. We do not have data on
per worker costs of HIV/AIDS prevention or pre-employment health checks but we
assume that these do not vary significantly across firms.
The Probit model used is as follows:
Where Y* represents the unobservable variable measuring the net benefit to a firm from
investing in any of these activities. The actual variable observed is y (whether or not a
firm carries out HIV/AIDS prevention or pre-employment health checks), measured as a
dummy variable, equal to 1 if Y* >0, and 0 otherwise. The function f is distribution
function--X is a vector of explanatory variables, and u is the unobserved error term.
The following equation is estimated for firm i, based on the simple model described
where Y = whether any HIV/AIDS prevention is carried out
=whether high-cost HIV/AIDS prevention activities are carried out
=whether the firm does pre-employment health checks
X1 = size of the firm, as measured by total number of workers
X2 = whether the firm is foreign owned (0/1)
X3 = ratio of skilled to total labor
X4 = whether or not a firm does training
X5 = whether or not a firm is majority unionize
X6 - X12 = sector and country dummies
The definitions of the dependent variables are as follows. The size of the firm is
measured by the total number of workers employed—part-time workers are assigned the
value 0.5. A firm is foreign- owned if it has greater than 10% foreign ownership. Skill
ratio is defined as the number of managers, professionals, and skilled production workers
as a proportion of total workers. Formal training is a dummy variable, equal to 1 if an
enterprise has a training program for its workers, zero otherwise. The majority unionized
dummy is set to 1 if more than 50 percent of workers in the enterprise belong to a union.
Sector dummies are assigned to firms in food processing, textile and garments, wood and
furniture, and metal working; sector effects for these key sectors are measured relative to
Unfortunately, we do not have data on the per-worker cost of HIV/AIDS prevention or
pre-employment health checks. However, it is unlikely that these vary substantially by
firm, for the simple activities we are considering (posters, condom distribution, VCT). In
alternative specifications of the above-described model, we also included a variable to
measure firm attrition as a measure of worker mobility. The rationale is that firms are
morely likely to invest in HIV/AIDS prevention if workers are less able to leave the firm.
This variable was not significant, largely because there is little variance in this measure
of worker mobility in our cross-sectional dataset. Presumably, panel data will be more
useful in this regard. Table 6 presents the results of the Probit estimations for firm
behavior. We estimate three econometric models that focus on (1) whether the firm
carries out a pre-employment health check, (2) whether the firm engages in HIV/AIDS
prevention, (2) whether the firm engages in counseling, testing (VCT)
Equation  presents the results examining the determinants of pre-employment health
checks.9 The dependent variable is defined as a dummy variable, equal to 1 if the firm
conducts a pre-employment check, zero otherwise. We see that firm size is extremely
significant- larger firms are much more likely to conduct pre-employment tests compared
to smaller firms. After controlling for firm size, we see that firms that provide have a
formal worker training program (beyond on-the-job) are much more likely to test new
workers. In addition, firms with a higher proportion of skilled workers are more likely to
engage in pre-employment checks.
It is interesting to note that after controlling for size, foreign ownership is not significant
in the multivariate estimation; foreign firms are not more likely to screen out potentially
sick applicants or carry out HIV/AIDS prevention.10 There is a fair bit of sectoral
variation as well, which may reflect differing degrees of labor mobility among other
things. Pre-employment health checks are significantly higher in the food sector, perhaps
for reasons of health and consumer safety. Finally, Kenya and Uganda do significant less
pre-employment health checks than Tanzania.
Equation  describes the results of the Probit estimation for whether or not the firm
engages in HIV/AIDS prevention activity consisting of posters, condom distribution
and/or VCT. We see that size matters here as well; larger firms tend to do more
prevention. After controlling for size, it is important to note that firms with better trained
workers and higher-skilled workers tend to do more prevention.11 Four sectors—food-
processing, wood, metal, and construction—tend to do more HIV/AIDS prevention than
other sectors.12 And the country dummies are not significant; there is no real variance in
9 Specifications that controlled for age of the firm and included a dummy for whether or not the firm is
credit constrained did not yield any additional significance.
10 The coefficient on foreign ownership is small and the variance is large, indicating that there is not
enough variance in foreign firms in our sample. Since most foreign firms are large, the size coefficient
captures the significance and foreign-ness alone does not give us additional information.
11 It could be argued that firms that do more AIDS prevention carry out more training i.e. that the causality
goes in the opposite directly. We do not believe that this is the case; anecdotal and other evidence suggests
that firms’s decision to do training greatly precedes their decision to do AIDS prevention.
12 It is likely the case that sectors where employees are not easily replaceable will do more AIDS
prevention; this may be due to issues such as lack of ease in hiring or the difficulty of losing workers
during peak seasons.
prevention activity across countries, after controlling for firm size and other firm
characteristics.13 The third column describes the determinants of more significant
HIV/AIDS intervention (Voluntary Counseling and Testing or VCT). Again, larger firms
and firms that have higher-skilled workers who are trained in-house tend to do more VCT
Why is size significant in the three regressions? Large firms may have better quality
managers, greater resources, and/or other unobserved characteristics that enable them to
do HIV/AIDS prevention.14 Larger firms may also have already-established facilities for
conferences or training that can be easily adapted for HIV/AIDS education sessions.
Apart from the fact that large firms may find HIV/AIDS interventions more affordable,
they may also be more aware of the risks of HIV. Available evidence suggests that small
and medium firms may be less aware of the risks of HIV, lack the staff and resources to
carry out prevention activity and are sometimes unaware of options available to them to
address the problem of HIV (Rosen et al, 2003, Durier, 2005 ). It is also worth noting
that foreign ownership is not significant after controlling for size. Finally, Kenyan firms
do more VCT activity than other firms, as do firms in the construction sector; the latter
perhaps because of the migratory nature of the workforce and/or the difficulty in
replacing workers in this sector.
Unionization is significant in determining HIV/AIDS prevention, only when a majority of
workers are unionized. A simple union dummy set to 1 if the firm has a union is not
significant, but a dummy recording whether more than 50 percent of workers are
unionized is significant in determining whether or not the firm carries out HIV/AIDS
prevention activity. Interestingly, it is also significant in determining whether or not the
firm carries out a pre-employment health check; this may be because firms with a
unionized workforce are aware that they have to provide a higher level of HIV/AIDS-
related services and may consequently do more to screen out sick workers. Other
13 This result is also reassuring in terms of the decision to pool the data across countries.
14 Anecdotal evidence suggests that large firms in the textile and garment sector in Lesotho carry out very
little AIDS prevention; this may be due to the highly mobile nature of firms in this sector. In countries
which serve as temporary homes to firms, one would expect less correlation with size or workforce quality.
econometric specifications that included measures of the regional and city location of the
firm, and age and education level of the manager, did not yield different results than
those reported here.15 Our results are quite robust to variations in specification.
Table 7 shows the probability of a firm of 66 employees (mean size for our sample)
carrying out pre-employment health checks and doing high-cost HIV/AIDS prevention,
based on the econometric results obtained in Table 7. We examine four scenarios for
each country, from a base case of a mean-size firm with a training program that is
majority-unionized, and has 35 percent of its workforce skilled.16 This firm has a 65
percent probability of doing a pre-employment health check in Tanzania. This number
drops to 54 percent if it does not have a training program or to 52 percent if it is not
majority-unionized. Thus, the impact of unionization and a training program are very
similar in terms of the likelihood of carrying out a pre-employment health check.
Increasing the firm size to 212 employees (one standard deviation larger) raises the
likelihood of a pre-employment check by 9 percent to 74 percent. The numbers for
Kenya and Uganda are also shown below. The results from this exercise confirm the data
reported in the descriptive tables in previous sections—larger firms, especially those with
higher investments in workers tend to do more HIV/AIDS prevention and to screen
employees more carefully; for these firms, it does appear that the benefits of HIV-related
activities outweigh the costs.
15 The lack of panel data prevents us from testing other hypotheses but we have made some attempts to do
address this issue in alternate econometric specifications. The addition of regional or city dummies did not
change our results ; these dummies were not statistically significant. One might argue that larger firms do
HIV prevention because of legal requirements. However, most legal requirements are de jure rather than
de facto in the East African context--analyses of the investment climate as well as of governance factors
have found that the regulatory and legal requirements are largely not binding (Kauffman et al, 2005).
Manager characteristics and experience with HIV may affect the decision to engage in HIV prevention or
pre-employment health checks--specifications of the econometric model that included manager’s age and
education did not show these variables to be significant, most likely due to lack of variation across our
sample of countries. Finally, there are not many stakeholder conflicts in the surveyed firms—the vast
majority are entrepreneur or family-owned on a privately-held, limited liability basis and less than 4
percent are publicly traded. We expect that these variables—legal requirements, managerial and
stakeholder attitudes—might be significant if a wider range of countries and geographic areas are
considered and if panel data are used. We are hopeful that panel data will be available in the future to
enable the investigation of HIV prevention activity in East Africa and elsewhere.
16 We calculate probability values from the cumulative density function underlying the numbers estimated
in Table 7.
IV: Worker Perceptions about HIV/AIDS
The dataset used in this analysis also contains information provided by workers within
the surveyed firms. A sample of 4950 workers was interviewed in East Africa, of which
80% of workers were male and 20% were female. Broken down by country, we have
1,922 workers interviewed in Kenya, 1,597 in Tanzania, and 1,436 in Uganda. Most
workers interviewed (84 percent) had permanent status, rest were temporary employees.
The majority of workers interviewed were between 20 and 40 years of age. The age,
occupational, and educational distribution of the workforce is shown in Table 8.
Unskilled production workers make up the largest share of the sample, followed by
skilled workers. We also see that about a third of workers have completed primary
school, another third have completed secondary school or vocational training, and 12
percent have a university degree.
The worker survey included questions about worker perceptions of HIV/AIDS. Workers
were asked to rank from 1-5 if HIV/AIDS was of concern to them. Close to 85 percent of
workers surveyed indicated that they are very concerned about HIV/AIDS, rating the
problem either 4 or 5 on the scale provided—there was little variation in the responses of
workers across age, occupational or educational status. Our data also show that about 75
percent of workers surveyed are willing to pay to be tested. This result is in sharp
contrast to anecdotal and case-study evidence which indicates that the uptake on free
testing provided by firms is very low.17 One explanation of our result is that workers are
telling the us what we want to hear i.e. they know that getting tested is “good for them”
and are consequently saying that they are willing to be tested. Another explanation is that
there is in fact a real interest in being tested but because of social stigmas or the visibility
of company clinics and VCT facilities, workers are reluctant to visit these health
facilities. If the second explanation is to be believed, there may be significant latent
17 A recent case-study provided by Debswana, a diamond mining company in Botswana, shows a VCT
uptake of about 20-25 percent.
demand for HIV testing.18 This high number may reflect, to some extent, the workers’
perception of the risk of being exposed to HIV.
Finally, Table 9 shows the amount that workers are willing to pay to get tested; we see
that there is a correlation between work status and amount that workers are willing to
pay. Interestingly, it appears that some workers are willing to pay an amount above the
actual cost of the test. If there is indeed latent demand, this might be realized if VCT
were part of a continuum of services, whereby workers have treatment options available
after learning their HIV status.
The analysis in this paper indicates that firms which are larger, have trained workers or
workers with greater skill levels, and/or are unionized do more to prevent HIV/AIDS.
These factors are also significant in determining whether firms do pre-employment health
Several questions emerge from this analysis--how we can create stronger incentives for
private sector intervention such as tax credits or other financial incentives? If larger
firms are doing more prevention as these results suggest, increasing their incentives to
provide HIV/AIDS prevention will increase the proportion of workers covered by some
prevention activity. If the result that larger firms do more is in part due to the lower
perceived benefit of AIDS interventions by smaller firms, can we raise awareness in the
small and medium enterprise sector about the true cost of HIV/AIDS? If size of the firms
drives the degree of intervention, the public sector will need to take the lead on
HIV/AIDS in most African countries for atleast the near future, given the high proportion
of small firms relative to large ones.
18 An informal discussion with Debswana staff was consistent with the second hypothesis; there is
considerable social stigma associated with being HIV-positive and the VCT service provided by the firm is
highly visible to all employees, perhaps explaining the low uptake.
We also need to find the means by which firms are motivated to do fewer pre-
employment health checks and more prevention activity. The screening process, as it
currently stands, is likely burdening the state as well as individuals and households by
imposing an additional constraint on income generation and the ability to deal with
If latent demand does indeed exist for HIV testing, both the public and private sectors
need to find ways to meet that demand--removing the social stigma attached to
HIV/AIDS testing and/or providing a continuum of services beyond VCT may be
necessary to ensure that workers are able to get tested. One implication is that if a fee-
based testing option were made available and all employees took the test, it would
become routine and might help end the stigma attached to testing since it would be a
market, consumer-oriented transaction.
The authors are grateful for detailed comments to four anonymous reviewers, and to
James Habyarimana, Elizabeth Ashbourne, Nancy Birdsall, George Clarke, Bill Cline,
Sabine Durier, Judy Feder, Alan Gelb, Alvaro Gonzalez, John Kline, Maureen Lewis,
Taye Mengistae, Agata Pawlowska, Axel Peuker, and seminar participants at the World
Bank, the Center for Global Development, and Georgetown University for helpful
comments and suggestions. The data for Sub-Saharan Africa used in this paper were
collected by the Regional Program on Enterprise Development in the Africa Private
Sector Unit of the World Bank. The views expressed in this paper are the authors’own
and do not necessarily reflect the views of the institutions with which they are affiliated.
Table 1: HIV Prevalence in East Africa in 2003
Adult (15-49) HIV
Adults and children (0-
49) with HIV/AIDS
Adults (Ages 15-49) with
Source: UNAIDS 2004; Center for HIV Information, University of California-San Diego
Table 2: Percentage of Firm Engaging in HIV/AIDS Prevention (weighted)
Note: Low is <5% population infected with HIV, medium is 5-10%, high is >10%
1.2 1.6 0.53
1.1 1.5 0.45
VCT (%) Both Types
Table 3: Prevention Activity by Firm Size in Kenya, Uganda and Tanzania
(% of firms in each size category, weighted)
Table 4: Percentage of Firms Conducting a Pre-Employment Health Check
(by country and region)
No Prevention Posters and
Table 5: Pre-Employment Health Check by Firm Size
(% of firms in each size category, weighted)
Table 6: Probit Estimations
* *** Significant at 1% level, ** Significant at 5% level,* Significant at 10% level. Standard
errors in parentheses. Skill ratio is defined as the number of managers, professionals, and skilled
production workers as a proportion of total workers. A firm is foreign- owned if it has greater
than 10% foreign ownership. Formal training is a dummy variable, equal to 1 if an enterprise
has a training program for its workers, zero otherwise. The majority unionized dummy is set to 1
if more than 50 percent of workers in the enterprise belong to a union.
- 3.44 ***
0.29 * *
Worker Skill Ratio
Textile and Garments - 0.14
Wood and Furniture
Table 7: Probability Values of Firms Doing Pre-employment Health Checks and
HIV/AIDS Prevention (%)
Firm Type Probability
In Tanzania In
Table 8: Age, Occupational and Educational Distribution of Workers
Age (years) 0-20 21-30 31-40
Occupation Managers Professionals Skilled
65% 18% 36% 26% 30% 25%
54% 13% 26% 13% 20% 19%
52% 10% 24% 10% 18% 14%
74% 35% 46% 46% 39% 44%
7.5 2 35 35
15 9 5 32
Primary Did not
12 33 33
Table 9: Willingness to Pay (U.S. dollars)
Means of maximum worker willingness to pay, by country and occupation
Health workers 8.11
Other non-production workers
(office, sales, service staff)
Skilled production workers 4.00
Unskilled production workers 3.39
Appendix 1: The Sample of Firms
Table A.1: The Sample of Firms in Kenya
Firm Size (%)Firm Location (%)
Small (10-49 employees)
Medium (50-99 employees)
Large (100-499 employees)
Very Large (500+ employees)
Market Orientation (%)Firm Activity (%)
Exporter (>=5% sales)
Chemicals and Paints
Paper, Printing, Publishing
Firm Ow nership (%)
Publicly listed company
Publicly held, limited company
Privately held, limited company
Foreign (>=10% foreign ownership)
Source: World Bank (2002), Investment Climate Assessment for Kenya
Table A.2: The Sample of Firms in Uganda
Firm Size (%)
Small (<100 employees)
Market Orientation (%)
Exporter (>= 10% sales)
Firm Ownership (%)
Publicly listed company
Publicly held, limited company
Privately held, limited company
Source: World Bank (2003), Investment Climate Assessment for Uganda
Firm Activity (%)
88.00 Chemicals & Paints
88.00 Paper, Printing, Publishing
Sample Textile & Leather
23.00 Central Region
2.00 North East
2.67 South West
Firm Location (%)
Table A.3: The Sample of Firms in Tanzania
Firm Size (%) Firm Activity (%)
Small (<100 employees) 74.28 Agribusiness
Market Orientation (%)
Exporter (>=10 % sales) 18.84 Paper, printing, publishing
Non-exporter 81.16 Plastics
Firm Ownership (%)
SampleDar es Salaam
Publicly listed 2.55 Arusha
Private held, limited 62.18 Mwanza/Mara
Publicly held, limited 2.18 Kilimanjaro
Cooperative 5.45 Tanga
Sole proprietorship 17.45 Kagera
Partnership 7.27 Morogoro
Government-owned 1.09 Iringa/Mbeya
Other 1.82 Zanzibar
Source: World Bank(2003), Investment Climate Assessment for Tanzania
25.72 Chemicals and paints
SampleFurniture and wood
Textiles, garments, leather 11.23
Firm Location (%)
Aurum Health (2005). Cost-benefit analysis of eight South African companies’
HIV/AIDS treatment programs, presented at the 2005 South Africa AIDS Conference,
Durban, June. http://www.aurumhealth.org
Aventin L and Huard P (2000). “The Cost of AIDS to Three Manufacturing Firms in
Cote d’Ivoire.” Journal of African Economies, 9(2): 161-88.
Biggs T and Shah, M (1997), “The Impact of the AIDS Epidemic on African Firms,”
RPED Working Paper no. 72.
Bloom D et al (2004), “Business and HIV/AIDS: Who Me? A Global Review of the
Business Response to HIV/AIDS,” Prepared for the World Economic Forum’s Global
Brink B (2005). Plenary session presentation on the HIV/AIDS program in Anglo-
American, 2005 South Africa AIDS Conference, Durban, June.
Connelly P and Rosen S. (2005), “Will small and medium firms provide HIV/AIDS
services to employees? An analysis of market demand,” South African Journal of
Coulibaly I (2005), “The impact of HIV/AIDS on the labor force in Sub-Saharan Africa:
a preliminary assessment,” International Labor Office, Geneva.
Debswana AIDS Office (2005), Presentation on Debswana HIV/AIDS Program,
Washington D.C., September.
De Walque D (2004), “How Does the Impact of an HIV/AIDS Information Campaign
Vary with Educational Attainment? Evidence from Rural Uganda,” World Bank Policy
Research Working Paper, No WPS3289.
Dickinson D and Stevens M (2004). “Understanding the Response of Large South
African Companies to HIV/AIDS.” WITS HIV/AIDS in the Workplace Symposium,
Center for Health, University of Witwatersrand.
Dorrington R, Bradshaw D, and Budlender D (2002). “HIV/AIDS Profile in the
Provinces of South Africa: Indicators for 2002.” MRC, ASSA, UCT Technical Report.
Durier S (2005). Meeting to discuss IFC Against AIDS Program, November 9.
Ellis Land Terwin J (2005). “The Impact of HIV/AIDS on Selected Business Sectors in
South Africa,” Bureau for Economic Research (BER), funded by South African Business
Coalition on HIV & AIDS (SABCOHA), October.
Fox M, Rosen S, MacLeod W, Wasunna M, Bii M, Foglia G, and Simon J (2004), “The
Impact of HIV/AIDS on Labour productivity in Kenya,” Tropical Medicine and
International Health, Vol. 9
Global Reporting Initiative (2005), www.globalreporting.org
Kaufmann D, Kraay A and Mastruzzi M, "Governance Matters IV: Governance
Indicators for 1996-2004," World Bank Policy Research Working Paper 3630, June
Lewis M (2005), “Addressing the Challenge of HIV/AIDS: Macroeconomic, Fiscal and
Institutional Issues,” Working Paper 58, Center for Global Development.
Maffessanti A (2006a). “HIV Absenteeism Study.” AIC Insurance Company, presented
at the 2006 International AIDS Conference, Toronto, August.
Maffessanti A (2006b). “The Impact of Absenteeism on Productivity at Welfit Oddy.”
AIC Insurance Company, presented at the 2006 International AIDS Conference, Toronto,
Mears B (2005), SABCOHA Presentation at South Africa AIDS Conference.
Morgan D, Mahe C, Mayanja B et al (2002), “HIV infection in rural Africa: is there a
difference in median time to AIDS and survival compared with that in industrialized
countries?,” AIDS, 16: 597-603
N’Galy B, Ryder RW, Kapita B et al. (1988), “Human immunodeficiency virus infection
among employees in an African hospital.” New England Journal of Medicine, 319:
Over M (1997). Confronting AIDS: Public Priorities in a Global Epidemic. World
Bank: Oxford University Press.
Panter C(2005). Presentations on Daimler Chrysler’s HIV/AIDS workplace program at
the 2005 South Africa AIDS Conference and 2006 International AIDS Conference
Rosen S (2001), “The Implications of HIV/AIDS for Nigerian Manufacturing Firms,”
RPED Working Paper No. 113, July.
Rosen S and Simon J (2003), “Shifting the burden: the private sector’s response to the
AIDS epidemic in Africa,” Bulletin of the World Health Organization 81(2): 131-37.
Rosen S et al (2003), “AIDS Is Your Business,” Harvard Business Review, Vol. 81, no.1,
Stevens M, Weiner R and Mapolisa S (2003). “AIDS and the Workplace: What are the
Managers in South Africa doing?” Presentation made at the South African HIV/AIDS
Conference, Durban, July.
Taylor K et al (2004), “Business and HIV/AIDS: A Global Snapshot,” Manuscript, July.
Turner G (2004). “Why do South African firms take action on HIV/AIDS?: Results
from 2001-2003 Survey Data.” Undergraduate Thesis, Stanford University.
Whittle H, Egboga A, Todd J et al (1992), “Clinical and laboratory predictors of survival
in Gambian patients with symptomatic HIV-1 or HIV-2 infection,” AIDS, 6:685-689
World Bank (2002), Investment Climate Assessment for Kenya, available at
World Bank(2003), Investment Climate Assessment for Tanzania, available at
World Bank (2003), Investment Climate Assessment for Uganda, available at