1 14 Agricultural Research and Produciiviiy 



The District Evidence 



Table 6.9 reports six sets of parameter estimates based on avail- 

 able data for the 140 districts. Two alternative dependent vari- 

 ables, the total factor productivity index and an index of 

 foodgrain yields per hectare, are utilized. The basic regressions 

 are regressions 1 and 4. We note that the state and regional 

 research variables are significant contributors to the statistical ex- 

 planation of both productivity change and foodgrain yields. The 

 state and regional research interaction variable is negative and 

 significant, thus confirming our expectations. The early period 

 productivity index has a significantly negative coefficient as ex- 

 pected, on the grounds that the higher the early period produc- 

 tivity gains, the lower economic slack at the beginning of the 

 period, and therefore the lower the potential for TFP gains in 

 future periods.'^ 



The lADP effect in regressions 1 and 4 is picked up by the 

 lADP dummy coefficient. It is positive in both cases. The statisti- 

 cal quality of the estimated effect is low in the case of regression 1 , 

 however. In regression 4, the estimated contribution to increased 

 foodgrain yields is highly significant both from a statistical and an 

 economic point of view. This is pretty much what should have 

 been expected of the program. By inducing producers to increase 

 the use of fertilizer and modern inputs, a large effect in yield 

 levels should have been forthcoming. As we have noted, 

 however, the real test of the contribution of the program is in 

 terms of productivity change. Our estimate shows this contribu- 

 tion to have been positive. 



Some supporters of lADP would argue that the real effect of 

 lADP is that it made research more effective. Regressions 2, 3, 5, 

 and 6 are designed to investigate whether the lADP had a strong 

 interaction with the research program. The state and regional 

 research variables are combined to form a new variable: 

 DISTR,, = kSR,, + LrR,, + LiSR X RR), 



13. An additional argumenl lor inclusion of the early period productivity gains 

 is that weather factors create a "regression" ellect that is partially controlled lor by 

 this variable. 1 1" beginning-period weather lactors are exceptionally lavorable, this 

 will lower the rate of productivity growth measured in following periods. It will 

 also be reflected in higher pre-IADP productivity growth. 



