The Case of India 101 



The regression estimates reported in table 6.5 show positive 

 and significant research coefficients, including a borrowing ele- 

 ment and a research-extension interaction effect. Drought and 

 region dummy variables have the expected sign. The northwest 

 region is the locus of the high-yielding wheat varieties. The posi- 

 tive dummy variable coefficient could be interpreted as the effect 

 of this exogenous technology. However, regressions similar to 

 those reported in the table were run for the period 1953 to 1965 

 with almost the same results. In all of the regressions (table 6.5), a 

 variable measuring the extent of irrigated acreage^ was included. 

 In all cases it was negative and insignificant. The widely held view 

 that the bulk of the productivity gains in India were associated 

 with irrigation investment and the Green Revolution does not 

 hold up. 



The Returns to Investment in Research 



The regression estimates (table 6.5) allow calculations of the 

 ''marginal" contribution of investment in research and in the ex- 

 tension system. Calculations based on the parameters estimated 

 in regressions 1, 2, and 3 in table 6.5, are shown in table 6.6. The 

 estimated contribution based on regression 3 is the "best" from 

 the point of view of the model. It has the following interpretation. 

 If the state research variable were to increase by 1,000 rupees, the 

 economic value of the higher level of productivity associated with 

 this increment would be 7,960 rupees (6,600 + 1,300). According 

 to the underlying model, productivity would remain at this higher 

 level, hence the 7,960 rupees is not a "once-for-alf gain, but 

 represents a "stream" of gains of 7,960 rupees, every year into the 

 future. There is, of course, a time lag between the investment and 

 the stream of returns. The lag specified in the model is of a dis- 

 tributive nature: it was assumed that a rupee spent in year t will 

 contribute to production over a ten-year period, the average lag 

 will thus be five years. It was estimated in a study of a similar lag 

 structure in the United States that the length of the average lag is 

 six-and-one-half years (Evenson 1971). Applying a six-and-one- 

 half-year lag to the 7,960/1,000 ratio yields an "internal" rate of 

 return of 40 percent to investment in state-based research. 



9. The ratio of wheat-irrigated acreage to net sown acreage. 



