Equation system (8) has been applied to a number of agricultural data sets (Huffman and Evenson 1993; Evenson et al., 1996) and technology variables have been incorporated into these systems. The 'shadow prices' (i.e. 9n*/9T) of technology variables can be computed and evaluated. Some of these systems have utilized MV variables as technology variables, although the exogeneity of such variables is in question. (Evenson et al, (1996) treat these as exogenous when computed for all crops.)
Equation system (8) does not utilize measures of crop acreage on the grounds that acreage allocation is implied by the supply decision and variable factor demand decisions. Of course it is true that the quantity supplied is simply acreage times yield, and we could justify replacing the supply equations in (8) with acreage equations and yield equations on these grounds. This would, by itself, be insufficient grounds for doing so, but there are at least four solid justifications. These are:
1. Acreage and yield decisions have a true sequential nature (Antle, 1983).
2. Error terms, especially weather errors, affect yields, but not acreage.
3. MV specifications can be endogenized.
4. Dynamic adjustment specifications can be more easily justified.
The sequential decision argument runs as follows. Farmers make their crop acreage decision based on information available prior to planting time. They make provisional input decisions at the same time. Once planting begins they cannot alter their acreage choice, but they can alter other input choices in response to changes in prices and weather events. Weather events may thus affect both factor use and yields.
Adjustment costs may impinge on the use of variable inputs as well as on acreage although one would expect acreage adjustment costs to reflect these variable input costs as well. (Family labour use may be a quasi-fixed factor with high adjustment costs, for example.)
Farmers respond to changes in technology as well as to changes in prices. Their ultimate objective is net revenue or net profits. They will compare net revenues from one crop with net revenues from another crop, then formulate expectations by observing MV availability and adoption as well as prices. The profits function model implies cross-equation restrictions on net revenues. Hence both MV or technology terms and prices will have these restrictions.
MV adoption itself should be treated as an endogenous choice variable. The logic of the traits discussion suggests that profitability and the availability of traits, along with farmer characteristics and extension, will govern MV adoption. One of the concerns in this specification is to measure trait availability so as to achieve 'exogeneity' for trait availability while allowing for endogeneity of the MV adoption itself.
In this study this is accomplished as follows:
1. MV profitability for rice is proxied by state acreage ratios of MV rice yields to traditional (unirrigated) rice yields. Dummy variables for districts are interacted with this variable to allow for proportional district differences. This variable reflects trait values to some extent.
2. For India, data have been collected for 'leading' rice varieties over the 1978-1992 period. In selected districts, farmers' yield traits for the three leading rice varieties were collected. The set of such varieties for each major agro-climatic region then constitutes a collection of ultimately successful varieties. For this set of varieties, it is possible through geneology analysis and breeders' ratings to compute acreage traits in the set of such varieties and to date them according to the date of release of the ultimately successful varieties. These 'availability' data are exogenous to farmers in that they represent breeders' success.
The model suggested by these considerations is shown as equation system
Was this article helpful?