Appendix 5: 

 The Research Variable 



Let Pi be the flow of knowledge created in year t. 



A stock and a flow (average) variables were specified and tried. 



t 



Stock ^(0= X) ^r 



where the lower bound of the sum is 1948. 



1 f 

 Flow (average) P(0 = 4; ^ /J 



r=t-4 



The stock variable in the 1955 analysis was the counts of publica- 

 tions over the period 1948-55, and the flow was the average of 

 counts over the period 1952-55. 



Publications were counted in the thirteen different sectors, 

 thus providing a measure of research work by sector. As research 

 work can be more important in terms of its contribution to pro- 

 ductivity, the larger the sector receiving the new knowledge, the 

 publications were weighted in the following way. Omitting time 

 and country indexes, define: 



q. Value of product in product and subsector / (wheat, 



maize . . .) 

 Q — X qj Total value of agricultural production in country 

 Pi Research (publications) in sector / 



P — 1, P- Research in the country. 



The weight used is Wj = q/Q. The share of subsector / in total 

 value of output and the research variable (flow) was calculated as 



13 



i? = 5^ p.w., i= wheat, . . . soil sciences 



Phytopathology and soil sciences were included in field crops 

 and weighted by the share of this category in total production. 



Weighted and unweighted, stock and flow research variables 

 were tested in regressions before the inclusion of a research vari- 

 able. Table 5. A reports two such tests. "Fto enter" is the F{t'^) 



\ll 



