70 A,i;riciilniral Research and Producllvily 



Area and time variables were always significant, though the 

 first varied in sign. The research and borrowing variables were al- 

 ways positive and significant in most regressions. 



All four estimates of the parameter in tables 4.2 and 4.3 indicate 

 intercepts [1/(1 + a:)] of the borrowing function whose values are 

 virtually zero, no borrowing taking place in the absence of in- 

 digenous research work. This somewhat surprising result was 

 confirmed in all the empirical experiments conducted.^ 



The economic implication of the estimates reported in tables 

 4.1-4.3 are discussed in the next section. But first a number of 

 "experiments'" should be described. 



1. Several regional classifications were tried. It was found that 

 the best results— in terms of R~ and of reasonable coefficients- 

 were achieved in wheat with the two-dimensional 3-4 classifica- 

 tion (appendix 3), and in maize with the five-dimensional 1-5 

 classification. These classifications were used in the analysis re- 

 ported in tables 4.2, 4.3 to construct the borrowing factor and the 

 regional deflators. 



The regional classification indicates the number of other coun- 

 tries that may have borrowed a unit of knowledge (one publica- 

 tion) produced in a particular country. The expected number of 

 borrowers of a publication is 5.32 in wheat and 1 .1 in maize, under 

 the assumption that a publication has the same probability of 

 being produced in any one of the countries in the sample (an 

 alternative assumption can be that this probability is proportional 

 to the current distribution of publications). This expected value is 

 termed here the transferability factor (appQnd'w 4). The values of 

 the transferability factors are slightly underestimated since not all 

 wheat- or maize-growing countries are included in our samples. 



2. Counts of articles from Field Crops Abstracts were tried in 

 wheat as an alternative to counts from Plant Breeding Abstracts 

 (averages of the two were also tried). Counts from Plant Breeding 

 Abstracts proved to be a superior variable. 



3. The element in the exponent of the borrowing function in 

 equation 4.7 is the flow of knowledge created. An alternative for- 



8. As the 0! values are not estimated at the sample means, they are subject to a 

 larger error than indicated by the regression results. 



