A cost-benefit analysis of female primary education as a means of reducing HIV/AIDS in Tanzania

Abstract

This article, uses panel data related to 20 Tanzania regions and 8 years to estimate the direct and indirect effects of female primary education on HIV/AIDS rates. A recursive framework for education, income and infections is employed, based on two autoregressive equations that allow us to obtain dynamic estimates of effectiveness. We find that the indirect effect working through changes in income outweighs the direct positive effect of education on infections, implying that female education can be effective as an intervention to lower the disease in Tanzania. The estimates of effectiveness are then utilized to carry out a cost-benefit analysis of the education expenditures. The human capital approach is used to measure the benefits. Irrespective of the timing of the benefits and costs, and the discount rate alternatives we consider, our best estimates result in positive net-benefits, with benefit-cost ratios in the range 1.3–2.9.

† This article was completed with the financial assistance provided by the Fulbright Research award for 2003. It was part of the project on the Cost-Benefit Analysis of HIV-AIDS interventions programs in Tanzania carried out when the author was a visiting Professor in the Department of Economics at the University of Dar Es Salaam in Tanzania.

Notes

† This article was completed with the financial assistance provided by the Fulbright Research award for 2003. It was part of the project on the Cost-Benefit Analysis of HIV-AIDS interventions programs in Tanzania carried out when the author was a visiting Professor in the Department of Economics at the University of Dar Es Salaam in Tanzania.

¹See Logie (1999).

²See Fawzi et al. (2004).

³For example, the Commission on Macroeconomics and Health (2001) called for the assistance from developed nations to Sub-Saharan Africa and other nations to rise from the current levels of US$ 6 billion to US$ 27 billion by 2007 and US$ 35 billion by 2015.

⁴See Household Budget Survey 2000/01 (2002), Table 4.2.

⁵One drawback of our blood donor data is that there exists large annual fluctuations in regional infection rates. However, this problem would not be removed if we relied on the ANC data instead as large annual fluctuations are present in this data source as well. For example, in the Iringi region (Regional Hospital site) the infection rate among clinic enrollees rose from 24.9% in 1998 to 40.1% in 2000 and then shrunk to 4.6% in 2000.

⁶It is illegal not to go to primary school in Tanzania.

⁷As we shall see in the empirical work, the only significant Z variables are time dummies and they are not functions of education.

⁸See Brent (2006).

⁹The World Bank (1999), p. 130, cites research for Brazil that shows that in 1985, about 3/4 of those newly diagnosed with AIDS had a secondary or university education, and by 1994 this proportion had decreased to 1/3. The World Bank predicted that ‘eventually’ the ‘positive relation in Africa’ would be reversed. But, this is not to say that this has already occurred, hence Brent's (2006a) positive results for 2001. The first research to find that the decline over time in Africa is such that it reverses in sign and becomes negative is by De Walque (2004) related to Uganda.

¹⁰The sign of α _Y is an empirical issue as it may not be negative. Income in Sub-Saharan Africa acts much like education, as Brent (2006) has found. It is the richest countries in Southern Africa that have the highest infection rates. In Tanzania, the level of income of a region was also positively related to the contemporaneous rate of infection. But, using the dynamic estimator we obtained the result that it is increases in income (not the levels themselves) that lower infection rates and we use this to produce a negative indirect effect.

¹¹If one matches the exogenous case for the infection equation (column 1 or 2) with the pre-determined case for the income equation (column 3 or 4) then the number of averted cases rises to 2297. It would be even higher at 4231 if one matches the exogenous case of the infection equation with the endogenous case for the income equation. However, we again point out the fact that the estimates for the education variable in the predetermined and endogenous cases in the income equation are not statistically significant using the one-step procedure, so we do not rely on these high numbers.

¹²See Morgan et al. (2002).

¹³Fifty years was the life expectancy at birth for 1988 stated in Table 13 of the poverty report by the Research and Analysis Working Group (2002).

¹⁴We start at 8 and end at 14, rather than at 7 and 13 (which is the legal primary school ages), because in practice primary school covers the age range 6 to 17 and we want to better represent this range.

¹⁵See Brent (2003), ch.7 on discount rate theory and practice in health care evaluations.

¹⁶The Household Budget Survey 2000/01 (2002), Table 6.3, gave the mean monthly education expenditure per capita as 227 TSHS in 2000/01 and 25 TSHS in 1991/2. We averaged these figures and multiplied by 12.

¹⁷The real growth rate in Tanzania over the period 1992 to 2000 was 3.1%, see Table 15 of the Research and Analysis Working Group (2002).

¹⁸The Household Budget Survey 2000/01 (2002), table B9.2, gives the monthly income for women with primary education as 19 990 TSHS, or 239 880 TSHS per annum. We adjust this amount upwards for each year after 2001 by the growth rate of 3.1% and then apply discount rates of 3% and 5% to obtain the net present value figures for earnings given in the text for profile 1. In the same table, monthly income for all women (with and without education) was slightly lower at 19 798 TSHS per month (237 880 TSHS per annum) and this is the earnings base for profile 2.

¹⁹For a comparison among CBA criteria, which includes a discussion of the NPV vs. the benefit-cost ratio, see Brent (1998), ch.2. As explained in the introduction, we do not assume that the funds for HIV-AIDS are fixed – an assumption which would make the benefit-cost ratio the appropriate criterion. We include both criteria outcomes for the reasons explained in the text (i.e. to facilitate comparisons with other evaluations).

²⁰For a comprehensive analysis of the strengths and weaknesses of the human capital approach relative to willingness to pay (WTP) in health care evaluations, see Brent (2003). While we basically endorse WTP as the superior methodology from the point of view of Welfare Economics, we need to point out that this approach assumes consumer sovereignty and that there are grave misgivings in applying this for expenditure decisions related to 7 to 14-year olds as in our study here. The human capital approach has the practical advantage of ready data availability. There are very few WTP CBAs for Africa. Examples are Forsythe et al.'s (2002) evaluation of voluntary HIV counselling and testing in Kenya, which used Contingency Valuation (CV) techniques, and Brent's (2006b) CBA of the condom Social Marketing Program in Tanzania, which relied on the revealed preference approach via estimated market demand curves. Primary schooling is now free in Tanzania (tuition before it was abolished in 2000 was about US $2 per year). So there is no scope to estimate market demand curves for primary education in Tanzania. There are no published CV studies of the WTP for education in Tanzania. CBA as applied Welfare Economics is covered in Brent (1996), which includes a number of CV applications (none related to Africa).

²¹We are using an exchange rate of 600 TSHS = 1 US$ for the monetary comparisons in this paragraph, the rate used by Sweat et al. (2000) in his evaluation of Voluntary Counseling and Testing in Tanzania.

²²The CBA of the condom socially marketing program in Tanzania by Brent (2006b) is an exception to this rule as it found that the program broke-even at existing subsidy levels, and it would have a benefit-cost ratio in the range 1.7–2.3 if the subsidy were reduced to its most efficient size. In this evaluation the condom costs were in US $ and the benefits were given by the area under the domestic demand curve for condoms.