The analysis of the optimal funding of education is complicated by the numerous and serious market failures which are likely to characterize a free market for education. Prominent amongst these are the likely external benefits of education, stressed by some new economic growth models. However, no model of education provision would be considered realistic if it did not also allow for: the central role of education in determining the distribution of income; the failures of secondary markets such as credit and insurance; the potential principal-agent problems and inequalities within the household; the merit good and limited consumer information, especially amongst uneducated parents; and the possible role of education as a signal of ability. All these market failures apart from the last tend to imply that a free market with no government intervention would underprovide education. Against these market failures, proponents of user fees for education have tended to stress failures apparent in existing government education systems. Fees are often advocated as a way of providing new funds for education in situations where financing from fiscally constrained governments is inadequate. The paper critically assesses these arguments, in the light of both the theory of segmented markets and the recent experience of developing governments in introducing or removing fees. There has been a long-standing and wide-spread international consensus in favour of re-structuring government expenditures from the tertiary to primary level. In part this is based on a misunderstanding: that the conventional (Mincerian) returns to schooling are higher for primary schooling than for post-primary schooling. Recent research (at least for Africa) has shown that this is not true of the 1980s and 1990s, and may not have been true prior to that. This does not fatally undermine the case for educational restructuring. However, at the very least it implies that the case must be centred more properly on arguments that the various market failures are more likely to occur at the primary than post-primary level. Such a case can be plausibly made, but as yet has not been fully substantiated empirically. Concern with government failure has extended to the provision, as well as financing, of education. Depending on whether the vouchers are limited to the state sector or not, such schemes could either create an "internal" market with the state sector or potentially be a form of indirect privatization. It is argued that whether vouchers have such dramatic effects will in part depend on how they are administered and in part on the relative cost-effectiveness of the state sector. In the short term, full blown unlimited voucher schemes may not appeal to many developing countries because their most direct effect will be a substantial increase in funding to users of private education (who are sometimes atypically affluent).