informal methods based on previous allocations; discus­sions and consensus among research managers taking into account national agricultural goals and strategies; to formal quantitative methods such as scoring models, congruency analysis, domestic resource cost ratio, mathematical pro­gramming, and simulation techniques. The more sophisti­cated approaches such as programming and simulation rely on mathematical optimization of a multiple goal objective function to select the optimum portfolio of AKST invest­ment. These are data and skill intensive and thus often quite costly to undertake. Many attempts have been made in the past to use a formal priority setting exercise to ensure that research resources are allocated in ways that are consistent with national and regional objectives and needs. Studies which have been undertaken to assess ex-ante AKST invest­ment priorities have included those employing criteria which include equity and distributive concerns (Fishel, 1971; Pin-strup-Andersen et al., 1976; Binswanger and Ryan, 1977; Oram and Bindlish, 1983; Pinero, 1984; Von Oppen and Ryan, 1985; ASARECA, 2005) those focusing more on effi­ciency criteria such as congruency (Scobie, 1984); those em­ploying the notion of comparative advantage using domestic resource cost analysis (Longmire and Winkelmann, 1985); those using economic surplus to examine research priorities (Schuh and Tollini, 1979; Norton and Davis, 1981; Rut-tan, 1982; Davis et al., 1987, Omamo et al., 2006); and those using an optimization routine (Pinstrup-Andersen and Franklin, 1977; Mutangadura, 1997). One of the most com­prehensive studies of research resource allocation lists meth­ods for allocating research resources (Alston et al., 1995). These combine information from scientists, technicians and other experts on the expected output of science, their prob­ability of success and possible timelines with information from economists and other social scientists on what the po­tential economic and social payoff would be if the research investment is successful. The formal methods have been extended to include environmental consequences of AKST investments (Crosson and Anderson, 1993). The overall aim is to foster consistency of research priorities with goals and objectives and to improve the efficiency of the AKST invest­ments in meeting the needs of the producers, consumers and society at large.
        In demand-oriented approaches, priorities are set based on the perspective of major stakeholders from outside the research system—especially the users. These might employ consultative and participatory methods using various forms of ranking techniques or users themselves might be empow­ered to make decisions on research priorities. However, it is worth keeping in mind that demand-led and supply-led approaches are not mutually exclusive. Better results can be obtained by combining formal supply-led priority setting with participatory approaches leading to better ownership of resulting priorities and greater chances that the priorities will be translated into actual resource allocation. Even the imperfect participation and empowerment of beneficiaries is likely to produce better results than conventional supply-led approaches on both efficiency and equity grounds, as they can improve the probability of broad-based adoption of technologies and knowledge generated, thereby enhancing innovation capacity. The challenge is to develop a judicious blend of bottom-up (demand-led) and top-down (supply-

led) approaches to priority setting incorporating the mul­tiple goals of AKST investments.
       Formal models exist for the ex-ante evaluation of re­search projects, which are being used increasingly in more industrialized countries to allocate research funds but this is less common in developing countries (Pardey et al., 2006a). Few formal ex-ante models incorporate the goals of reduc­ing poverty and hunger and the environmental consequences as explicit criteria for allocating research resources. Some progress has been made recently to incorporate these aspects in the analytical process. The two ex-ante studies reported in Eastern and Southern Africa (ASARECA, 2005; Omamo et al., 2006) consider the ex-ante benefits of all major com­modities and the economic and poverty reduction poten­tial of research investments. In addition, there are specific studies on site-specific maize research in Kenya (Mills et al., 1996) and the research priority setting under multiple ob­jectives for Zimbabwe (Mutungadura, 1997; Mutungadura and Norton, 1999). The extent to which such results are actually used for setting the R&D agenda remains unclear. These approaches (based on expected costs and benefits) are very useful in allocating resources among applied and adap­tive research programs and projects. However, they are of very little use to allocate resources between basic, strategic, applied and disciplinary research.
       It is not just methods per se that are problematic; it is also the ability of would-be analysts gaining the requisite skills to use what methods are available. In the context of NARS, the task of developing the needed capacity to address aspects such as environmental and economic assessment of agricultural technology consequences on NRM (Crosson and Anderson, 1993) is still not yet adequately developed, especially in an era of profound underfunding of research, at local, national and regional levels. An important issue in developing and implementing AKST investment priorities is to explicitly incorporate the requirements of those who are expected to benefit from such investments.
        Our approach in this study, which presents the empiri­cal evidence available on the economic, health and environ­mental impacts of research but does not try to use a formal priority setting process to weight the importance of different criteria, reflects the discussion of well-intentioned, but often misguided attempts to deal with such multi-criteria formula­tions of research priorities (Alston et al., 1995). The review of methods based on scoring models suggests that there are definitely methodological challenges in such work yet to be satisfactorily dealt with. This fact shows the need of more resources to develop easier and more effective evalu­ation methods that can include environmental and societal (poverty, nutrition and health) impacts, both positive and negative.

8.4.2 Investment options
The ideal social planner would be able to rank research in­vestments by their expected contribution to economically sustainable development, decreased hunger and poverty, improved nutrition and health, and environmental sustain-ability; and then would solicit weights from society based on the relative value society places on these expected contri­butions. Each country will have different weights based on the governance of the system and the countries' available re-