Funding agricultural RD in developing countries

514 | IAASTD Global Report • There is no measurable difference in estimated ROR be- tween privately and publicly performed research; • The RORs were 25 per year higher for research on ield crops and 95 per year lower for research on nat- ural resources than for total agriculture; • There is no signiicant difference in rates of return relat- ed to whether studies reported basic or other categories of research; • The estimate also indicates that if research took place in an industrialized country, the ROR was higher by 13 per year, but this effect was not statistically signiicant at the 10 level. The estimated rates of return tended to be lower in Africa and West Asia and North Africa than in Latin America and the Caribbean or Asia; • There is no evidence that the ROR to agricultural RD has declined over time; • Unable to detect any effect of accounting for spillovers or market distortions on measured rates of return to research.

8.2.4.4 Agricultural research and education investments and agricultural growth

A summary of studies that have applied decomposition analysis to agricultural growth in developing countries sug- gests that past investments in agricultural research may have contributed anywhere from 5 to 65 of agricultural rates of return. The highest ROR observed for all agricul- ture, ield crops, livestock, tree crops, resources and forestry were 1,219; 1,720; 5,645; 1,736; 457; and 457, respectively. All studies related to livestock and trees had a positive ROR. The mean ROR for livestock RD was around 121. These data demonstrate that the estimated RORs for livestock spe- cies are comparable to the rates estimated for the other sec- tors. In addition, in this study the overall estimated ROR for animal research was 18 but when this was decomposed, the ROR for animal health research and animal improve- ment research were found to be 15 and 27, respectively; indicating the underestimation of ROR for the overall in- vestment. Probably, the decomposition by species would also show different RORs associated to each of them. Although the mean ROR estimates for industrialized countries is higher than that for developing countries 98 and 60, respectively, the median are virtually identical 46 versus 43 Table 8-12. While there are not many studies from Africa assessing the returns to RD, the exist- ing analyses generally indicate high returns in the range of 4 to 100 for country level studies Anandajayasekeram and Rukuni, 1999. The key indings of the last meta-analysis were Alston et al., 2000a: • Research has much higher ROR than extension only or both research and extension combined; Table 8-10. Ranges of rates of return. Sample Number of observations Rate of return Mean Mode Median Minimum Maximum count percent Full sample a Research only 1,144 99.6 46.0 48.0 -7.4 5,645 Extension only 80 84.6 47.0 62.9 636 Research and extension 628 47.6 28.0 37.0 -100.0 430 All observations 1,852 81.3 40.0 44.3 -100.0 5,645 Regression sample b Research only 598 79.6 26.0 49.0 -7.4 910 Extension only 18 80.1 91.0 58.4 1.3 350 Research and extension 512 46.6 28.0 36.0 -100.0 430 All observations 1,128 64.6 28.0 42.0 -100.0 910 a The original full sample included 292 publications reporting 1,886 observations. Of these, 9 publications were dropped because, rather than speciic rates of return, they reported results such as 100 or 0. As a result of these exclusions, 32 observations were lost. Of the remaining 1,854, two observations were dropped as extreme and inluential outliers. These two estimates were 724,323 and 455,290 per year. b Excludes outliers and observations that could not be used in the regression owing to incomplete information on explanatory variables. Source: Alston et al., 2000a.