The Case of India 1 1 5 



where ^4, ^5, and o^are the estimated coefficients from regression 

 1. DISTRjjlhQn measures the estimated contribution of all 

 research to TFP in district / at time /. By multiplying this by the 

 lADP dummy for lADP districts we get 



DDISTR.^ = (DUADF)X DISTR.^ 

 and for non-//4Df districts 



NDISTR.^ = (1 -DIADP)X IDSTR.^ 



We then estimate the following equation: 



TFP.^ = C+ b^DDISTR.^ + b^NDISTR.^ 



+ bj^TFP566\. + b^^DDR.^ + b^^DREG. + e(6.5) 



The coefficients b^ and bg test whether research affected the 

 lADP districts and non-IADP districts in a different way. They 

 allow us to test whether the slope coefficient on the research vari- 

 able differs in the lADP districts. Regression 2 indicates that the 

 marginal contribution of research toward increased productivity 

 is not higher in lADP districts. Regression 5 has yield as the de- 

 pendent variable, and it indicates that the marginal contribution 

 of research toward increased yields is greater in the lADP dis- 

 tricts. Regressions 3 and 6 add the lADP dummy variable, allow- 

 ing both the intercept and slope terms to differ for the lADP dis- 

 tricts. We find that the slope coefficient in regression 3 is greater 

 for non-IADP districts, i.e. the marginal contribution of research 

 to non-IADP districts is greater than to lADP districts, while the 

 opposite is true for yields. 



These relationships indicate that lADP programs complemented 

 the research inducement to increased yields, but substituted for 

 research in terms of the contribution to total factor productivity. 

 That is, it increased the marginal contribution or "product" of 

 research toward increasing yields, but decreased the marginal 

 contribution of research to total factor productivity. This result is 

 quite plausible since many of the lADP activities would be ex- 

 pected to substitute for research. 



