Those aspects of tsetse and trypanosome biology and ecology that militate against sustained suppression of the vector and disease are quite well-understood (e.g. Leak 1999); enduring difficulties in sustaining control initiatives are largely institutional (Swallow and Woudyalew 1994; Barrett and Okali 1999; Bauer and Snow 1999; Politzar et al. 1999). Figure 1 reveals that these institutional constraints derive in the first instance from the technological difficulties raised by the non-linear relationship between tsetse challenge and trypanosomosis prevalence. All difficulties faced in collective community-based tsetse control (Swallow and Woudyalew 1994; Barrett and Okali 1999), and more generally in integrated vector and disease control (Tamboura et al. 1988; d'Ieteren et al. 1999) derive from this relationship.
But non-linear relationships such as that depicted in Figure 1 also imply major economic opportunities. For the resulting indivisibilities not only erect barriers to entry, production and trade, they also entail increasing returns to scale, favouring expanded economic activity. The greater is an organisation's ability to override the technological indivisibilities linked to an activity, the higher should be its capacity to enter into that activity, and thus the brighter its prospects for capturing associated scale economies, and vice versa. However, as illustrated by the Ghibe example, some of these economies may accrue to others as externalities, dampening private incentives to produce them.
While the lumpiness and indivisibility of public goods—such as tsetse control in Ghibe—may be long-recognised (e.g. Ellickson 1973), less well-appreciated is that, at the level of an economic system (i.e. spanning production, exchange and consumption), most private goods also exhibit important indivisibilities that may militate against efficient purely private production and delivery.
Consider inorganic fertiliser, an archetypical private good. Significant indivisibilities and scale economies arise from lumpy, spatially-dispersed, and seasonally-explicit demand for the input, alongside high domestic transportation costs, and intense financing and knowledge requirements in international trade. Together, these factors produce sizeable start-up costs for trade in fertiliser, rendering domestic markets imperfectly competitive at various levels and thus, by definition, constrained by the demand-side. Fertiliser traders gravitate toward locations with potentially high demand. But even there, limited quantities are brought to market at high prices. These high prices depress demand for fertiliser as producers economise on its use, many by pursuing diversified, subsistence-oriented production strategies. Returns to such practices are low. This further dampens demand for fertiliser, which lowers incentives and returns to increased procurement and distribution by traders, save for at a higher price. The result is a self-reinforcing movement toward an equilibrium featuring limited trade in high-priced fertiliser alongside a preponderance of production systems featuring limited use of the input, low productivity, and low returns. Absent an intervention that addressed one or more of the system's inherent indivisibilities, no supply-side or demand-side pressures will appear to reverse the outcome.
The significance of the Ghibe trial therefore lies in its success in overriding a set of inherently indivisible physical and institutional obstacles to growth—a set that is rife in African agriculture (IFPRI/ISNAR 1991).
Which matters more between physical and institutional impediments to growth is an empirical issue that depends on specific circumstances (Demsetz 2000). Indeed, they are likely jointly determined (Gotsch 1972). ILRI's most important interventions focused on the institutional constraints. Specifically, its efforts to raise awareness among farmers, help set up treatment centres, and organise treatment of sick animals produced a crucial combination of effects: learning effects (which acted to lower the uncertainty surrounding pour-ons as their pervasiveness increased), co-ordination effects (which conferred advantages to 'going along' with other villagers taking similar action), and adaptive expectations (where increased prevalence of pour-on use in the region enhanced beliefs of further prevalence). These effects led to insecticide adoption rates and use-intensities sufficient to overcome the inherent lumpiness of tsetse control (captured in Figure 1) and catalyse the self-reinforcing process described earlier. Absent this role—which sprung from a concern for the kinds of collective, regional effects that typically motivate public bodies but with which few private operators would have been concerned—the intervention would almost certainly have failed.
A defining characteristic of agricultural activity is that it is dispersed widely over space. By implication, so, too, are agricultural incomes. That spatial distribution of income appears to have been fundamental to the success of the Ghibe initiative. It mattered because pour-ons produced their collective effect in the 'wake' of their privately consumable effects. This suggests that not only must the precise nature of extension services affected by reform be ascertained—pure public versus mixed private–public versus pure private (e.g. Umali et al. 1994)—so, too, should be the extant distribution of rural income. For few 'public' goods are purely public; collective effects are typically accompanied by private gains and losses (Bergstrom and Goodman 1973; Ellickson 1973). Incomes of consumers and users will matter. And for rural goods and services consumed widely over space, the distribution of those incomes over the landscape will be crucial.
Another key aspect of ILRI's role in the Ghibe experiment thus was that it led to an 'organised articulation of demand' that resolved co-ordination problems typically involved in 'market creation' (Edquist et al. 2000). In so doing, it quite quickly formed a 'core' of innovative users sufficiently large, and sufficiently widely dispersed over space, to dissolve the 'trap' (of high vector challenge and high disease prevalence, and thus low human and livestock population, and thus high vector challenge and high disease prevalence, and so on) in which most tsetse-infested regions find themselves. In a very real sense, the strategy employed (and the process that emerged as a consequence) was the exact reverse of those in privatisation and decentralisation programmes currently dominating extension reform in Africa, which implicitly assume that the principal constraints on improved agricultural technology development and diffusion are on the supply-side (e.g. Rivera et al. 2000; World Bank 2000). The internal logic of the Ghibe trial was also fundamentally different from that driving the 'demand-led farmer-participatory' approach to extension reform, which assumes away the question of how a well-articulated demand for improved technologies and new information appears to begin with.
The primary lesson for extension reformers emerging from the Ghibe example thus is that it is the demand-side of an agricultural extension system that matters the most–i.e. the demand for new information, which itself derives from demands for such goods and services as improved inputs and credit. In effect, the question of how to 'get extension right' reduces to how to spur sustained and concentrated demand for new agricultural knowledge and information–i.e. demand for information about new cultural practices, new inputs, new crops and pests, new traders, new lenders and conditions in newly relevant distant markets.
This suggests both bad news and good news for efforts to reform agricultural extension in Africa. The bad news is that appearance of these various demands is far from automatic. Where, as in much of rural Africa, infrastructure is poorly developed and production environments are risky, natural forces will likely continue to define economic outcomes, muting demand for productivity-enhancing innovations, and translating into limited and costly supply. The good news is that if the jointly determined physical and institutional indivisibilities that these natural forces produce are properly identified and addressed, non-linear jumps in demand for innovations can occur, leading to rapid and sustained increases in productivity growth and welfare. The paradox facing agricultural extension reform in Africa thus is that its problems and solutions rest on isomorphic processes. This paradox drives to the core of the challenge for extension reform in Africa.
Most public agricultural extension systems in Africa are virtually declining. Increasingly more is being made of production and delivery of agricultural knowledge and information through private, market-based channels, on one hand (e.g. Rivera et al. 2000; World Bank 2000), and through local, collective, community-based channels, on the other (e.g. Farrington 1998). But the current analysis points to poor prospects for both strategies viewed, as they often are, in isolation. For in the Ghibe trial, ILRI—through its interest in the efficacy, impacts, and sustainability of the pour-on technology—devised an institutional solution to a deep biophysical nonlinearity and in so doing catalysed a multi-faceted and self-affirming movement toward low vector challenge and disease prevalence. Market-led and community-based extension reforms will succeed only if they contain similar solutions to inherently lumpy problems of rural development on which public sectors have stumbled. There is nothing spontaneous about such solutions. They must be carefully designed and implemented. More strongly, the Ghibe example indicates that the question of how to reform agricultural extension in Africa should not be framed as a mutually exclusive choice among markets, local community-based collective action, and public sector investments. Rather, the challenge lies in how to appropriately interface these three options in technology development and diffusion. The example illustrates, first, that the central ingredient in that interface is concentrated demand for new knowledge and inputs, and, second, that the physical and institutional factors that impede 'natural' appearance of that demand are also those that will unleash it.
