Southern African Regional Poverty Network (SARPN) SARPN thematic photo
Country analysis > Tanzania Last update: 2019-05-21  

Global Poverty Research Group

The dynamics of returns to education in Kenyan and Tanzanian manufacturing1

Mеns Sцderbom2, Francis Teal2, Anthony Wambugu3 and Godius Kahyarara24

Global Poverty Research Group

April 2005

SARPN acknowledges the ESRC Global Poverty Research Group as a source of this document:
[Download complete version - 544Kb ~ 3 min (39 pages)]     [ Share with a friend  ]


Returns to education remain of central policy concern in both developed and developing countries. In developed countries observed rises in returns to education have been imputed to skill biased technical change (Katz and Autor, 1999). In poorer countries such as those of Sub-Saharan Africa (SSA) it has been argued that returns may have been falling as a result of rapid expansion of education. As educational supply grows without a commensurate rise in demand, the probability of getting a job for any given level of education declines and, among those with jobs, returns may fall.5 The limited evidence of how wages have changed in Africa in the recent past suggests they have fallen, possibly substantially. Squire and Suthiwart-Narueput (1997) document that real minimum wages halved in Kenya between 1970 and 1985, and fell even more in Ghana and Zaire. However it is known that in many countries minimum wages are not enforced and the numbers are uninformative as to how the returns to education have been affected.

Policy interest focuses not only on the average return to education but on the dispersion of returns across education levels. A prominent feature of policy towards education in SSA has been the priority given to expanding primary education (e.g. World Bank, 1995). Such an emphasis is seen as being justified, in part, by the finding that returns to education are highest at lower levels (Psacharopoulos, 1994; Psacharopoulos and Patrinos, 2001). The shape of the earnings function is a key factor for understanding how policies of education expansion will impact on incomes. If innovations in educational policy impact primarily on those with high education costs, and the earnings function is concave, then returns to such reforms will be relatively high. However, the view that the earnings function is concave in education has recently been challenged for both developed and developing countries.6 If in fact the earnings function is convex, so that the marginal returns to education are lowest for the individuals with the least education, giving priority to investment in primary education may have little impact on poverty unless the individuals affected by the reforms proceed to higher levels of education.7

For poor developing countries, such as those in SSA, evidence is very limited as to what the earnings function looks like (its shape) and how the returns to education have changed over time.8 This paper considers these issues using comparable repeated crosssection data on workers in manufacturing firms in Kenya and Tanzania over the 1990s. We also put the results in a longer term context as excellent data exist for the returns to education in Kenya and Tanzania in 1980 (Knight and Sabot, 1990). We thus fill a significant information gap. Further, the different educational policies pursued by Kenya and Tanzania in the period since independence have been argued to constitute close to a natural experiment (Knight and Sabot, 1990). In the 1980s, while Kenya allowed a rapid expansion of secondary education, much of it privately financed, Tanzania severely restricted access to secondary education and introduced wage polices to reduce differentials. In the 1990s educational and other policies in Tanzania became much more similar to those in Kenya. A comparative analysis of these two countries over this period will therefore shed light on some of the general connections between education policy, education and earnings (e.g. whether the returns to education change when policies change, and if so, how quickly).

The rest of the paper is organised as follows: Section 2 outlines our empirical framework; Section 3 discusses the data and shows summary statistics; Section 4 shows OLS estimates of the earnings functions, and provides a comparative analysis over time and across age groups; Section 5 shows additional results in which education is treated as an endogenous variable; and Section 6 provides conclusions.

  1. We are grateful to Jonathan Temple (Editor), two anonymous referees, Marcel Fafchamps, Geeta Kingdon, Jim Malcomson, Janvier Nkurunziza, Margaret Stevens and participants at a CSAE lunchtime seminar and a Labour Economics seminar at the Department of Economics, Oxford, for several constructive comments on earlier versions of the paper. All errors are our own. Correspondence to Sцderbom at
  2. Centre for the Study of African Economies, Department of Economics, University of Oxford, UK.
  3. Department of Economics, Kenyatta University, Nariobi, Kenya.
  4. Economic and Social Research Foundation, Dar es Salaam, Tanzania.
  5. Bennell has recently argued as follows: "During the 1960s and 1970s, obtaining a 'good' job in the rapidly expanding formal sector of the economy provided a powerful incentive for households to invest in primary education (especially for boys). During the last 20 years, however, wage employment opportunities have contracted sharply in many countries as have formal sector incomes especially in the public sector", Bennell (2002, p.1186). An influential argument that educational expansion will be self-defeating in that it will simply result in higher qualifications being needed for any given job can be found in Dore (1976). The implications of his argument for Kenya and Tanzania in the 1990s can be found in Toyoda (1997) and Cooksey and Riedmiller (1997).
  6. In a series of papers Bennell (1996a,b; 2002) has argued that the pattern of the returns to education do not follow that asserted by Psacharopoulos (1994). Bennell's underlying arguments are consistent with the shape being convex. Direct evidence of convexity in some parts of the domain of the earnings function is provided by Belzil and Hansen (2002) for the U.S. and by Kingdon and Unni (2001) and Duraisamy (2002) for India.
  7. Throughout the paper we consider the Mincerian returns to education which do not reflect the private costs, other than foregone wages, or the possible non-wage benefits. It may be that the social returns to primary education are high e.g. in terms of health, but we are unable to investigate this with our data. Further, because primary schooling is a necessary input into postprimary, the prospect of postprimary schooling may raise the primary return above the rate as conventionally measured (Appleton, Hoddinott and Knight, 1996).
  8. Much of the available evidence is limited to relatively short periods of time. Krishnan, Sellassie and Dercon (1998) show that educational returns did not change in urban Ethiopia despite labour market reforms instituted in early 1990s. In contrast in Uganda, from 1992 to 1999, returns to education increased markedly, Appleton (2002) and in Ghana from 1987 to 1991 there is evidence of rising returns, Canagarajah and Thomas (1997). Where longer run comparisons have been made there is evidence of falls. In South Africa, Moll (1996) reports that returns to primary education declined from 1960 to 1975. In urban labour markets in Kenya between 1978 and 1995 Appleton, Bigsten and Manda, (1999) report declines in returns to education for workers with secondary education and below.

Octoplus Information Solutions Top of page | Home | Contact SARPN | Disclaimer