The conflict between the wish of educators in Africa to supply vocational education while students and their parents demand an academic one has been a continuing theme in discussions of educational policy in Africa since the now classic works of Foster (1965a,b). Tanzania is a country which has been through a cycle of policy making which began in the 1960s with an attempt to shift the educational system towards a more vocational focus and a restriction on the supply of secondary education (see Psacharopoulos and Loxley (1985) and Knight and Sabot (1990)) and ended in the 1990s with a reversal of many of its key educational policies. While Tanzania's shifts have been more dramatic than most a general trend away from vocational schooling to more general academic education was apparent in the 1990s. This process had powerful backing from those investing in education. The Education Sector Review of the World Bank (1995) argued that the rate of return was much higher to investments in general than in vocational secondary education.
This 'new' view - that it is academic not vocational training that should underlie educational policy - has not gone unchallenged. Bennell (1996 a, b) reports some higher rates of returns to vocational education than the rates of return to general education and argues strongly against any underlying presumption that academic education has a higher return than vocational.
A similar argument can be found in Bennell and Sergerstrom (1998). Two papers which report higher returns from vocational than academic education are a study by Neuman and Ziderman (1989) for Israel and one by Moenjak and Worswick (2003) for Thailand.
Indeed research findings do appear to be inconclusive. Zymelman (1976) in a review of school-based vocational training concluded that there was no clear evidence either for or against this type of educational provision. Chung (1995) in a review of the literature undertaken from the 1970s to the early 1990s found that 12 studies on returns to vocational education in developing countries reported higher returns to vocational training, 10 studies reported lower returns to vocational education or not different from other forms of learning, and 5 studies concluded that there is no basis to compare the returns to vocational education with the returns from other forms of learning. Other studies have stressed that the returns to vocational education depend substantially on the general level of economic development, the availability of private sector jobs, and whether or not people are employed in a field related to their training (Bennell and Segerstrom, 1998; Middleton et al., 1993).
In parallel with the concern to promote skills within schools by means of vocational schooling has been the perceived need to promote training within firms to address problems of poor productivity. A large literature has developed arguing that limited skills is the key to understanding poor performance in African countries, particularly in their manufacturing firms, Pack (2002). The central premise of much of this discussion has been that markets for skills will not operate and that there is a need to subsidise firms to ensure that the training occurs.
In this paper we ask three questions flowing from these concerns. First, what can account for the continuing strong preference for academic education in Africa where the level of development is so low and wage jobs are expanding so slowly (see Kingdon, Sandefur and Teal (2005) for a review of the empirical evidence for this assertion)? Secondly, what can account for
the diversity of the findings in the literature regarding the returns from vocational and academic education and is any general answer possible as to which has the higher returns? Thirdly, and most generally, which forms of educational investment are most profitable in terms of increasing incomes - vocational school, technical college, academic education or on-the-job training? We
will argue that the answers to all these questions are linked through the shape of the earnings function and the role of firm effects in determining earnings.
In the next section we set out the background as to how enrolment has changed in Tanzanian schools over the period from the 1960s to 2000. In section 3 we set out the earnings function we will use. An extensive literature has been concerned with two econometric problems that arise in estimating such functions in developing countries. The first is the possibility of a selectivity bias as wage earners are not a random sample of the population, Moenjak and Worswick (2003) find a much higher return on vocational education for Thailand when they allow for selectivity. The second problem is that the return on education may be biased up if ability is omitted from the equation. Both these issues are extensively discussed in SС†derbom et al (2006) for the data that will be used in this paper. They can find no evidence from instrumenting that the returns from education go down, which is in line with virtually all research in this area (see Card (2001) for a review). In this paper we follow their control function
approach of using the residuals from an auxiliary for education to test if the results are biased due to the endogeneity of education. In section 4 we set out how we propose to use the data to address the issue of how returns from vocational and academic schools can be compared. Sections 5 reports results for the sample as a whole, section 6 sets out the returns by the size class of firms. In section 7 we assess the role of firm fixed effects and worker quality in determining our results. We return to the most general issue under review - the returns to vocational relative to academic education - in section 8. A final section concludes.