ecuted optimally (Evenson, 2001). An AKST program may have other relevant impacts, such as social (poverty reduc­tion, enhanced nutrition, equity) and environmental effects; and benefits of the research may be distributed in different ways. Some nonmarket impacts such as environmental or health effects of AKST could potentially be given economic value and incorporated into economic analysis. Measure­ment in these cases is, however, usually more difficult than the measurement of economic impacts that are observable in product or input markets. These attributes should be ac­counted for in some way, even if economic values cannot be ascertained, when a more realistic evaluation of research impacts is required. In any meaningful empirical analysis, a multi-criteria approach is recommended to assess the im­pact of AKST assessment
.            The literature on economic impact studies includes a wide range of levels of impact analysis. The economic basis for government involvement in agricultural AKST is the per­ception of market failure leading to private underinvestment (Nelson, 1959; Arrow, 1962; Alston and Pardey, 1998). The appropriate criterion for the assessment of policy aiming to correct market failure is the effect on net social benefits, and this can be expressed as a social rate of return (ROR) to public investment in agricultural AKST (Alston et al., 2000a).6

In the literature the terms financial, economic and social rates of returns mean different things, but in this chapter the term economic rate of return and social rates of return are used in­terchangeably. This is because the various meta-analyses do not explicitly make this distinction.

8.2.3 Methodological limitations of impact measurements

Although there have been significant developments in im­pact assessment methodologies a number of issues still need further attention. Key among these are the issues of attribu­tion, incrementality, causality, defining counterfactual situ­ations, and estimating economic impacts for organizational and institutional innovation and social science and policy research. The issue of counterfactual situations refers to the significant problem of determining what the pattern of pro­ductivity growth would have been in the absence of a par­ticular research investment (Alston and Pardey, 2001). This is associated with dynamics of productivity factors even in the absence of AKST investment. AKST programs operate in environments in which ordinary or "natural" sequences of events influence outcomes. Impact assessment and ROR estimates must arrive at estimates of net intervention effects, i.e., they should measure the changes attributable to the in­tervention.
         Causality is another issue that merits attention. In mea­suring the impacts of AKST investments, it is important to ensure that the impacts measured are the results of the technologies and activities undertaken within the program/ project. However, as one moves from the direct product/ output to broader economic, social and environmental ef­fects, the chain of causal events is too long and complex, and the variables affecting ultimate outcomes are too numerous to permit the identification and measurement of impacts of specific interventions (Biggs, 1990; Rossi and Freeman, 1993). This is further complicated by the time lag between initial investment and reaping its return.
        Attribution problems arise when one believes or is try­ing to claim that a program has resulted in certain outcomes,