Research and Piodiiciiviiy in Wheat and Maize: An International Analysis 65 



serve as a deflator. However, regions can vary a great deal in size, 

 so the second deflator used was 



where rij'xs the number of regions in the country and Vj = l/«; 

 the average regional share. Thus. 



-^ = — + F. (4.10') 



2/ ; 



where P^-, like the variance, is a measure of dispersion. The inclu- 

 sion of t^- corrects for unequal size distribution of regions. 



Estimates 



Regression specifications and estimates are presented and 

 preliminarily discussed in this section; economic implications are 

 brought forward in the next. 



Two principal sets of estimates were calculated: (a) cross-sec- 

 tional, ''rate" regressions (table 4.1); (b) a combination of time 

 series and cross-sections, ''yield'' regressions, (tables 4.2, 4.3). 



The estimated weighted'' "rate" equation (table 4.1) was 



p. = b K.(6S) + cp^. K(68) + u. (4.1 1) 



7. Unweighted regressions and weighting by area were also tried. The following 

 table gives R^ values for regressions 1 and 3 of table 4.1 



Weighting did not improve the estimates of the yield regressions. We are indebted 

 to Finis Welch for the suggestion that we use the average square deviations as 

 weights. 



