Research and Productivity in Wheat and Maize: An International Analysis 69 



research in the later period. A possible explanation for this find- 

 ing can be the complexity of modern maize technology. Rapid 

 yield improvement in the past may be an indicator of the tech- 

 nological maturity of the country's agricultural industry — a prere- 

 quisite for the successful absorption of sophisticated modern in- 

 novations. 



As noted earlier, the contribution of research activity will be 

 overestimated due to the omission of other yield-affecting factors. 

 However, to the extent that publications measure research output 

 with random error, the regression estimates are biased downward. 

 The specification in table 4.1 may introduce another source of 

 random error. In the table, A^(68) stands for the stream of 

 knowledge created, which affected yields over the twenty-one 

 year period. This formulation implicitly assumes similar time pat- 

 terns of research work in the different countries, which was not 

 the case. 



A more detailed analysis becomes feasible with the utilization 

 of a combination of time-series and cross-sectional data. The 

 regressions estimated were of the general form 



^r=To+7i4v+7/+73(^y,/^) 



+ \iB.Jd)+y^S.+u.^ (4.13) 



The regressions were calculated in linear and in double-log 

 (Cobb-Douglas) form. In the double-log form all variables except 

 time were replaced by their natural logarithms, and one was add- 

 ed to all stocks, K, B, to avoid zeros. 



Regression estimates for wheat and maize are reported in tables 

 4.2 and 4.3, respectively. The country-specific variable, Sj, con- 

 tributes importantly to R} in these regressions. In the absence of 

 this variable, R? is of the order of .4. Results of this nature should 

 be expected when countries differ substantially in their yield po- 

 tentials (see columns 1, 6 in appendix 2). The inclusion of this 

 variable converts the regression into a covariance analysis, the 

 coefficients measuring the "within" country effects. This elimi- 

 nates biases that may have been introduced by correlation be- 

 tween knowledge, own or borrowed, and yield-level potential. 



